I am connected through sshonto a GPU and running a CNN keras model training with tensorflow-GPU within a Jupyter Notebook.
I left the process running and when I came back ColdTurkey had mistakenly closed the browser window in my laptop (but the process was still running in the Terminal) so I openned a new window again, and I saw that the relevant jupyter cell has the little square empty ([] rather than [*] which shows not running / running). However, the kernel shows busy, and when I run the command nvidia-smi on the Terminal, the Volatile GPU-Util is being used (and I am the only user which appears connected to the GPU).
I have tried openning and closing the browser link several times with no success. When I initially run the cell with the training, I left a couple of cells waiting in the queue to be run which are supposed to save the model and loss graph. So, if it was the case that the processes will finish eventually, I am happy to leave it running in the shawdows until nvidia-smi shows no Volatile-GPU Usage which should indicate it is done. However, I really need this processes to run so I do not want to wait unnecessarily if the processes are corrupt somehow and I am not expecting it to run properly. I would guess the first option is the case here but I would really appreciate if someone could explain to me what could be wrong with the browser visualisation.
If it helps, this is my terminal output:
