Own trained nn with weights

I am really struggling to deploy my own neural network. I have trained models on my own computer and get all the necessary weights. But I have no idea how to deploy it on my jetson nano. When I tried to launch this just from IDE, it is extremely slow. Who faced with the same problem and how did you solve it?



Which frameworks do you use? PyTorch? TensorFlow?

Here is our tutorial for deploying a model on Jetson for your reference:


Hi! Thanks for your replay.
I am using PyTorch. Thanks for the link, probably I missed something but I did not find the information that I need. It only said how to train your model, but I have trained this already and just want to use it on Jetson Nano.


You can just use pyTorch to run your model.
Please check this topic for the installation steps for pyTorch:

On Jetson, we also have an accelerated engine for deep inferencing called TensorRT.
If you want to convert your model into TensorRT, please check this GitHub for information:


Well, I follow the instruction to install PyTorch (It took 24 hours!!!). But I still do not know what is the problem. When I launch the model, the camera starts working and even show me one frame and next but very very slow.
And Windows where the camera is open become black and white. (This is a link on image https://ibb.co/p2HrDH4).


Have you maximized the Nano performance first?

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

A possible issue is that your model is too complicated to run on Nano.
To check this, it’s recommended to monitor the system status at the same time:

$ sudo tegrastats

Please pay attention to the memory and CPU/GPU usage.