Your HuggingFace model, fine-tuned on local GPUs.
What it does
Builds and runs a reproducible HuggingFace fine-tuning pipeline on local NVIDIA GPUs, with optional Hub publishing.
When to use it
Use for HuggingFace model fine-tuning, training pipelines, or local smoke tests when no dedicated TAO model skill owns the model.
How to use it
Give it a HuggingFace model ID; it generates a training project and runs data preparation, smoke testing, fine-tuning, evaluation, and inference.
What you provide
- a HuggingFace model ID
Uses
Good to know
- cannot be undone: HF Hub
- edits its own instructions
- acts without asking
Latest review
What you need · 10
Docker must already be available; the initial probes run in a CPU-only container.
Requires NVIDIA Container Toolkit for GPU-enabled Docker execution.
Access · 7
This skill
HF Hub
Write
Publishes your fine-tuned model (irreversible)
WandB
Write
Logs training to WandB