tao-train-deformable-detr - Train and run TAO Deformable DETR
Train, evaluate, export, quantize, and run inference for TAO Deformable DETR 2D object detection models.
Tags
Updated: 2026-09-28Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Train Deformable DETR models
- Evaluate object detection models
- Run model inference
- Export models to ONNX
- Quantize model weights
- Generate TensorRT engines
- Run AutoML training
Inputs
- COCO or COCO-raw datasets
- Image archives
- Annotation JSON files
- Class map files
- Model checkpoints
- Action specifications
- Spec overrides
- Dataset class metadata
Outputs
- Training checkpoints
- Evaluation metrics
- Inference predictions
- Exported ONNX models
- Quantized model artifacts
- TensorRT engine files
- Action status and logs
Requirements
- Docker
- NVIDIA Container Toolkit
- TAO Deploy container for TensorRT generation
- Packaged train schema and template for runnable AutoML