Installing TAO Deploy 5.5.0 on Jetson Orin Nano (JetPack 6.1/5.1)

Hi all,

Goal
I want to run Mask2Former’s instance-segmentation model on a Jetson Orin Nano via a TensorRT engine. I’ve already trained the model on an x86_64 machine and exported it to an ONNX file.

What I’ve tried

1. Followed the TAO Deploy Jetson instructions for version 5.0.0—Mask2Former isn’t included until 5.5.0.
2. Used L4T TensorRT Docker images r8.6.2 and r10.3.0, but neither provides TAO Deploy 5.5.0 via pip.

Environment

• Device: Jetson Orin Nano
• JetPack: 6.1 (also tried 5.1)

Questions

1. What’s the recommended way to install TAO Deploy 5.5.0 on Jetson Orin Nano?
2. Are there offline or wheel-based methods if it’s not on PyPI yet?
3. Do I need to add NVIDIA’s PyIndex or use a different repository URL?

Any tips or pointers would be greatly appreciated—thanks!

1 post - 1 participant

Read full topic