Get your GPU host ready for TAO workloads.
Checks GPU hosts for TAO compatibility and, with user approval, installs or repairs the required NVIDIA runtime stack.
When to use it
Use before running TAO workflows on Docker, local-Docker, or Kubernetes GPU hosts.
Give it a GPU host setup or compatibility-check request; it reports missing components and repairs the host only after approval.
Checks support any Linux distribution; automated installation supports the documented Debian, RHEL, and SUSE families.
Targets NVIDIA GPU hosts and requires an x86_64 or aarch64 (sbsa) host.
This skill
/etc/apt/sources.list.d/nvidia-container-toolkit.list
Replaces the NVIDIA package source (irreversible)
Kubernetes
Checks cluster GPU capacity