Hello NVIDIA supporter,
We are using both L4T R35 and R32 image docker to develope ourr application. We found that the in L4T R35 jetpack, everything is download and packaged (cuda, cudnn, tensorrt…) in 1 docker so that the final jetpack image has a big size (10GB). However in R32, the cuda, cudnn libraries are copied from the host volume , so that the L4t R32 image is just consumer a small size.
We wonder whether have a guide for us to customize the l4t jetpack image docker by copying the lib from the host volumes for R35 similar as R32?