Tao Train Dino

DINO is NVIDIA's transformer-based object detector with denoising anchor boxes, wrapped for TAO workflows. You get train, eval, export, distill, quantize, and inference actions, all backed by proper dataclass schemas and AutoML support out of the box. The skill enforces a validation dataset even for training runs because DINO unconditionally builds a val dataloader, so you'll need both train and val URIs or you can reuse the same path for both. It ships with a small synthetic COCO subset for smoke testing (4 object classes), and the reference docs cover the full TensorRT deploy path. The AutoML policy defaults to on, and the skill routes through tao-run-automl unless you explicitly turn it off.

skills.sh installs: 1.6K
GitHub stars: 3.9K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add nvidia/skills --skill tao-train-dino --agent claude-code

Installs into .claude/skills of the current project.

Files

DINO

Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery).

DINO (DETR with Improved DeNoising Anchor Boxes) for 2D object detection. Transformer-based detector with denoising training, multi-scale features, and optional distillation support.

Uses pretrained backbone weights (e.g. ResNet-50 ImageNet). Set model.pretrained_backbone_path for backbone-only or train.pretrained_model_path for full model.

When To Use

Train, evaluate, export, distill, quantize, or run inference for a TAO DINO 2D object detector.

For TAO Deploy TensorRT actions (gen_trt_engine, TensorRT evaluate, and TensorRT inference), read references/tao-deploy-dino.md first. Deploy spec templates live in this skill's references/ folder with the spec_template_deploy_*.yaml prefix.

Reference Map

SKILL.md on GitHub

Details

Repository: github.com/nvidia/skills
First seen: Jul 14, 2026