Closed pgrandinetti closed 6 years ago
Entering the e-mail and token at the end was designed to resolve this issue. The idea is,
the scenario I imagine is
Last I looked into it the token timeout was stated to be around 30 mins (although I have not observed this in practice), which is too short for the assignments where each problem could take a few hours to get 10 points. Also note that the solutions of all problem parts need to be submitted together to get full credit. This is a requirement of the Coursera platform. You might need to run your solver for 6 hours locally and then perform the submission.
Oh, that's it. I thought the token verification could be done at any moment (for instance, as first step in the submit.py script). If that's not the case than this ticket can be closed :)
Unfortunately not, it all has to be sent to Coursera as a single payload see, https://github.com/discreteoptimization/assignment/blob/master/submit.py#L326
You might give the _credentials
file a try, but this will fail if Cousera expires your assignment token.
I see. Then maybe you could count the total time from the moment the first instance starts till the moment the last instance is completed, and if that is < 10h then let the program waiting for email+token even if they are entered much later?
I don't follow your last comment. The code does all of the computations then it prompts for email+token. It's fine if they are entered much later, as long as the token is still valid in Coursera's system.
If the computation is going to be some hours long then I'd rather prefer to enter my email and submission token before to start the computational part, otherwise I need to be sure I come back at the computer to enter email+token before the 5h/10h deadline expires. And if I run it overnight that may be a bit annoying.