#deployment

Shipping a model to the edge: PyTorch → ONNX → TensorRT

A hands-on, current (2026) path for taking a PyTorch CV model to ONNX and a TensorRT engine: export, parity, FP16/INT8 build, and latency gating.