Predicting Generalization in Deep Learning Forum

Go back to competition Back to thread list Post in this thread

> Submissions Taking Extremely Long

Hello!

Could you please help me with re-submitting submissions. I'm just trying to submit the baselines, and it's taken hours or just been stuck on the 'submitting' phase. Anything I can do to make things go faster or work better?

Thank you so much!

Posted by: anican @ Aug. 18, 2020, 9:52 a.m.

I have the same issue, the code has been blocked in the 'submitting' status for a long time. Could you please help us to fix the problem? Than you!

Posted by: Ethan_Udel @ Aug. 18, 2020, 1:53 p.m.

Hi, if a submission is stuck at submitted. It means that there are not enough servers to run the submission, and it is the expected behavior.
If something is stuck in "running", then there could be a bug.

Posted by: ydjiang @ Aug. 19, 2020, 11:36 p.m.

Thanks! Is there a way we can test if there's a bug before we submit? What are the requirements for submission again? In our zip file we must have an __init__.py, metadata file, and complexity.py with a complexity function? Am I missing something? I tried to make a zip of the baselines_2 (sharpness), and it fails

Posted by: anican @ Aug. 20, 2020, 9:11 a.m.

Apologies that sharpness_v2 has some known bugs. I will try to correct that as soon as possible. Otherwise, you can check this post (https://competitions.codalab.org/forums/21997/4223/) for the fix.
Every other baseline should run fine locally. The README shoudl contain information about how to test submissions locally. You can also use the provide Colab to test your submissions.

Posted by: ydjiang @ Aug. 23, 2020, 1:01 a.m.

Hi all,

I have a submission that is "running" since 09/18/2020 09:43:40, could you stop it please?

Best

Posted by: cadurosar @ Oct. 6, 2020, 12:43 p.m.

Hi,

If a submission is stucked for more than 24 hours it means it has already timed out (hard expired on the workers).
In case like this, sometimes Codalab fails to update its UI.

Posted by: ydjiang @ Oct. 6, 2020, 9:48 p.m.
Post in this thread