A working training stack for your DGX Spark.
What it does
Prepare a working ML training or inference environment on DGX Spark with compatible containers and packages.
When to use it
Use for fresh environment setup, import or framework installation failures, container-versus-pip decisions, or restoring an environment after system updates.
How to use it
Give it an environment setup or troubleshooting request; it prepares a container-first training or inference environment on the computer.
Uses
What you need · 2
Targets DGX Spark hardware with a GB10 Grace Blackwell chip, aarch64 CPU, SM121 GPU, 128GB unified memory, and CUDA 13.
The default container path needs Docker configured to use the NVIDIA runtime and expose the GPU.
Access · 3
This skill
GitHub
Checks recent GitHub issues
docker
Starts a GPU-enabled container