Hello,
We submitted a model for Task 1 and received the error: Container did not generate a file called predictions.csv. We reviewed the logs provided after the submission but didn’t find any trace of what could be the issue. Prior to submitting we tested in different machines and the codes were able to generate the output/predictions.csv file. One possibility we are considering is whether this could be due to a memory issue, if assuming the input .h5ad file used for run may require more memory to load/process. Would you have any suggestions for how we might debug further?
Thank you.
Created by Iliza Nazeraj iliza.nazeraj Thank you for the confirmation. Hello @iliza.nazeraj, Yes, I can confirm that Cognitive Status is only available in the training dataset due to its correlation with the metrics that are being predicted. All other variables/data (except those that are being predicted) are available. -Kyle Hi @iliza.nazeraj ,
Sure thing! The container logs had the following error:
```
Error in `select()`:
! Can't select columns that don't exist.
x Column `Cognitive Status` doesn't exist.
Backtrace:
... [R backtrace truncated]
Execution halted
```
I am re-confirming with the organizers on whether the "Cognitive Status" is only available in the training dataset.
Hello,
We made a submission with the id: 9760032 and received the error: _Container did not generate a file called predictions.csv_. Are there any additional logs we can check to help troubleshoot the issue?
Thank you. Thank you! @iliza.nazeraj ,
Thank you for the extra context. We have updated the resource constraints for the containers, and your submission will now be re-run. You should be receiving updates soon! Thank you for your reply.
We tested with input the .h5ad files provided and it succeeded at ~256GB but failed when limited to 32 GB. @iliza.nazeraj ,
Thanks for reaching out! You are right - based on your submission's Docker logs, an error was encountered while loading the input data. We will discuss internally and if needed, will re-run your submission.
For reference, how much memory did your model require when you were running locally?