I want to deploy one of tlt nvidia model on jetson nx, but the image of jetapck 4.4 requried at least 16 GB space in memory.
Q1- I want to know, in your opinion, Is it possible to reduce size of image to only support the model?
Q2- In the jetpack 4.4 there are some package suck as vulkan/openGL/NSight, … There are accessory?
Q3- Is it possible to use docker container for this? Is this way can reduce the image size?
Q4- 16 GB jetpack image in my opinion is high, I want to know, Which of libs or packges used high space?
Really we need whole of this image in production?
Q5- I guess the huge part of this image is for OS, and I disable GUI and gdm and other services, but these are have small effect in image size, I want to know Is it possible to use light OS for this and then I installed some of packages like cuda/cudnn/TensorRT, …? Is it possible to control the size of image with SDK manager flashing into jetson?