Search. Pull. Build.

NVIDIA Data Center GPU Manager (DCGM) is a project for gathering telemetry and measuring the health of NVIDIA GPUs

Minimal gpu-feature-discovery container image.

gpu-fryer is a Hugging Face tool that stress-tests NVIDIA GPUs and flags abnormal thermal throttling or performance degradation, which is especially useful for finding the slowest GPU in a multi-GPU system.

The NVIDIA GPU Operator bootstraps, configures, and manages GPUs in Kubernetes.