NVIDIA-AI-IOT/torch2trt

Health v2 · maintenance only

Adoption not part of the score

4878 stars · 698 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

torch2trt is a Python library that converts PyTorch models to TensorRT engines using the TensorRT Python API, with a simple single-function conversion interface and extensible custom layer converters. It was created primarily to optimize models for NVIDIA Jetson embedded platforms.

Use cases

• convert pytorch models to tensorrt for faster inference
• optimize deep learning models on jetson nano or xavier
• speed up resnet or alexnet inference with tensorrt
• run pytorch classification models faster on nvidia gpus
• save and load tensorrt-optimized models as state dicts
• write custom tensorrt layer converters for unsupported ops

When to choose

• you deploy PyTorch models on NVIDIA Jetson devices and want easy TensorRT acceleration
• you want a minimal, single-call converter for common vision models
• you need to extend conversion with custom Python-registered layer converters

When to avoid

• you need full coverage of all PyTorch ops or production-grade conversion - use torch-tensorrt or ONNX export
• you are not on NVIDIA hardware with TensorRT available
• your model uses many unsupported layers that you cannot write converters for

Facets

library · maturity maintenance

machine-learning llm-inference compiler gpu-computing deep-learning gpu-computing developer-tools python pytorch tensorrt model-conversion inference-optimization jetson nvidia linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NVIDIA-AI-IOT/torch2trt")