I have a large dataset (that I am trying to run. The cell has not generated an output; however, it currently says 'completed at [time]'. The cell still appears to be running and the is a message saying 'waiting for python 3 to Compute Engine Backend.'
does anyone know whether the the cell timed out? should I rerun, or should I leave it as is?
Some of my runs have also encountered exactly the same conditions as yours. Then, after acknowledging some sort of implementation that enforces interactive use from Google (reCAPTCHA - which was added earlier this year, pops up randomly during your runs, if you fail to click it, your runtime will be terminated within minutes), I come up with the following theory.
My guesses are that, when you are unable to click the captchas (lucky case where it doesn't get terminated), or you left the VM running for too long without any interactions with it (the latter has much higher chances), Colab may decide to silently terminate your runtime shortly. If no additional interactions are seen within some time later, it will perform the full termination (your code and results will be gone).
I have observed some of the times where I left it running and switched to another tab, when I switched back, it performed a reconnection and then showed the "completed at [time]" but the code was still running. Other times I observed no reconnection but the postconditions were the same as yours. The training results and other data on VM were left intact regardless. So, keep running it will cause no problems in my experience, no reruns are needed, although you should keep an eye for your VM when the condition occurs to avoid any possible problems.
