Failed algorithm while tryouts are working well

Dear organizers,

We are a bit lost..

Our Submissions for the validation phase do not go through, however the tryouts on the same algorithm work with no problem, in low time.

Could you help us understand what is making our submissions fail?

Regards

Zhack

Dear Zhack,

Your submissions fail because the algorithm takes longer than the 5-minute limit per case for this challenge. This issue likely arises because the test files are larger than your sample file. The algorithm runs out of time during what appears to be test-time augmentation, but it seems your preprocessing is taking too long (approximately 3 minutes). You can review the logs here: https://grand-challenge.org/algorithms/autopetiii_chb/jobs/.

In the Data-centric tutorial, we address this issue by using test-time augmentation (TTA) that adjusts based on the remaining time, allowing smaller files to receive more TTA than larger ones.

Best Jakob

Thank you,

For the record, we do not use TTA, which means something is definitely wrong ^^'

We will explore this

Thanks again

Zhack

We still have access to the run results, is that on purpose ?

Zhack

Yes, the preliminary files are just data samples from the training set. This makes it easier for participants to debug their code. Each team should only see their own results.

Best Jakob