Support and FAQ

Use this page as a quick starting point when you need help with NVIDIA NIM for Large Language Models. It highlights the most useful pages in this documentation set and answers common operational questions.

For the most common tasks, refer to the following pages:

• Support Matrix for supported models, deployment profiles, and verified hardware.

• Prerequisites for software, licensing, and NGC API key requirements.

• Quickstart for a minimal local deployment.

• Model Download for NGC, Hugging Face, S3, GCS, and local model download workflows.

• Architecture for container startup flow, health endpoints, and request routing.

• Logging and Observability for logs, metrics, JSON Lines output, and tracing.

• Helm on Kubernetes for standard Kubernetes deployments.

• Cloud Service Provider Deployment for AWS, Azure, Google Cloud, and Oracle deployment guides.

FAQ

How Do I Confirm That My System Is Supported?

Start with the Support Matrix and Prerequisites. These pages document the supported models, validated deployment profiles, required GPU configurations, and minimum software versions.

The Container Starts, but Readiness Stays Unhealthy. What Should I Check?

Refer to Architecture to understand the startup sequence. A container can be live before the model is fully loaded, so /v1/health/ready can return 503 during model download or initialization. Then refer to Logging and Observability for log and metrics collection, and Model Profiles and Selection if you suspect a profile or GPU resource mismatch.

How Do I Authenticate Model Downloads?

Use an NGC Personal API key for NGC-hosted containers and models. Refer to Prerequisites for API key requirements, Configuration for export examples, and Model Download for source-specific download behavior.

Which Deployment Guide Should I Use?

Use Quickstart for a simple single-container deployment. Use Helm on Kubernetes for standard Kubernetes environments, OpenShift for Red Hat OpenShift, Multi-Node Deployment for large distributed models, and Cloud Service Provider Deployment for managed cloud environments.