README.md

USLS Examples

This directory contains practical examples for all models supported by usls. Each demo shows how to configure, run, and integrate state-of-the-art vision and vision-language models.

🚀 How to Run Example

Execute a model demo (e.g., RF-DETR) with hardware acceleration:

# Model + processor on CUDA
cargo run -F cuda-full --example object-detection rfdetr --device cuda:0

# CUDA model + CPU processor
cargo run -F cuda --example object-detection rfdetr --device cuda:0 --processor-device cpu

# TensorRT + CUDA processor
cargo run -F tensorrt-full --example object-detection rfdetr --device tensorrt:0

# TensorRT + CPU processor
cargo run -F tensorrt --example object-detection rfdetr --device tensorrt:0 --processor-device cpu

Always use --release (or -r) for production.

---

CLI Arguments

Argument	Description	Example
--source	Input media source	path/to/img.jpg
-p, --prompts	Text prompts	-p "cat"
--device	Inference device	cpu, cuda:0, tensorrt:0
--processor-device	Pre-processing device	cpu, cuda:0
--dtype	Model precision	fp32, fp16, q8
--batch	Batch size	1, 4

Run cargo run --example <name> -- --help for model-specific options.