GPU autoscaling on Kubernetes with KEDA: building an external scaler with NVML

If you run vLLM, Triton, or any other inference server on Kubernetes, you have probably noticed that...

Raccontata dadev.to