I am quite a newbie in this field. I have access to a remote server (Debian 10) that I can access via ssh. The server is equipped with a Nvidia GPU.
I installed the CUDA version 11.3 and now I realized that the latest tensorflow only supports CUDA 11.2.
Removing everything is so painful, so I followed another guide from tensorflow website (https://www.tensorflow.org/install/docker):
- I installed docker
- I installed nvidia-docker
- I pulled
docker pull tensorflow/tensorflow:latest-gpu-jupyter
But now, I stuck somehow. What I need are:
- Be able to start jupyter on the remote server, and access that from my local machine (the remote server does have an external static IP).
- Be able to install new packages (say pytorch, sklearn, etc.) along tensorflow.
- Be able to save the data in the hard drive of the remote server so I can reuse it.
I have no idea how to do any of the above. Hope that you can give me some guidelines.
Many thanks