Hi,
For TX2, please take a look at documents in download center:
https://developer.nvidia.com/embedded/downloads#?search=tx2
And we suggest use DeepStream SDK for running deep learning inference. Please take a look at
Announcing DeepStream 5.0.1
Announcing new software update for DeepStream 5.1
NVIDIA Metropolis Documentation
If you have a TX2 developer kit, please try deepstream-app.