How to setup specific version environment to run Deep Learning Application on the JetsonTX2 without re-flashing the Jetson board again?

I would like to run a GitHub repository on my Jetson TX2. Packages requirement to run the application on the Jetson TX2 is following:

  1. Cuda 9.0
  2. OpenCV 3.4.1
  3. Caffe 1.0.0
  4. Pytorch 1.0.0
  5. Python 3.6

Packages version available currently on my JetsonTX2 are:

  1. Python 3.7
  2. CUDA 10.2
  3. PyTorch 1.0.0

I tried use the following docker command (“nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04”) to pull the docker image container; but, docker has deprecated the support for NVIDIA CUDA (mentioned on the docker website).

  1. Is there any way to install all the packages with the specific version at once?
  2. Also, versions of this packages are old so most of them are archived so how can I install the archived of the packages?
  3. Do I need to re flash the Jetson TX2 to setup this packages version environment?