Need Detailed Submission Failure Logs

Competitions AI for Industry Challenge
1

Hi everyone, our team name is starvla.

Our model runs correctly in the local environment, but the submission has failed several times after uploading.

Our setup uses micromamba to launch a separate inference environment/process for the VLA model, and the model checkpoints/parameters are packaged together with the submitted image.

What we observed:

  • The reported execution time on the submission platform is around 800s before failure

  • In our local environment, the same pipeline finishes successfully in around 2 minutes

  • Inference works normally locally

  • There are no available logs after failure

Currently, we can only see the header information containing the team name and submitted image ID, but no actual runtime logs.

Could any administrator provide more information about possible submission restrictions, timeout limits, or ways to access detailed logs?

Thanks!

2

We have already tried moving large imports/model loading into __init__() and other runtime functions to avoid time exceeding issues.

We also removed the original time.sleep logic by using step-skipping methods to avoid problems.

However, the submission still fails without any available logs.

3

See also the thread here, that’s got a wealth of discussion around it. Many of us have this problem. Error Outputs on Submissions

4

I’ve noticed this thread. Thank you for your reply. If you have any extra findings, please let me know.