Launch training runs and find out why they stall.
What it does
Launches, monitors, stops, and diagnoses NeMo-RL training runs on Kubernetes using nrl-k8s.
When to use it
Use when asked to launch, iterate on, or debug a NeMo-RL recipe on a Kubernetes cluster.
How to use it
Give it a training launch, monitoring, or debugging request in the project; it returns shell commands, configuration instructions, and analysis.
What you provide
- An existing project
Uses
KU
Good to know
- cannot be undone: Kubernetes
Latest review
What you need · 9
Uses the nrl-k8s CLI to validate recipe and infrastructure files and submit training runs.
Requires kubectl to verify cluster state and inspect resources before acting.
Access · 3
This skill
Kubernetes
Write
Creates and deletes cluster resources (irreversible)
RayCluster
Execute
Starts clusters for training