Grading queue backlog since the v0.10.0 rollout

Hi,

Thanks for the update and deadline extension. I am seeing a significant grading queue backlog after the v0.10.0 rollout.

Across 14 submissions today:

  • 6 submitted around 00:00 UTC were graded only after roughly 5 hours in ready;
  • 3 submitted around 05:50 UTC are still ready after more than 3 hours;
  • 3 submitted at 07:43 UTC are still ready;
  • 2 submitted at 08:41 UTC are still ready.

Eight submissions are currently queued, and none from the last two batches has started grading.

This may be caused by the daily limit reset, the ongoing re-grading, and increased submission activity after repricing. Since submissions now run on one core, each job may also occupy workers for longer.

With Phase 1 closing on 10 August, multi-hour turnaround makes it difficult to compare variants reliably.

Would it be possible to allocate additional compute capacity and/or speed up the grading process while the current backlog is being cleared?

Thanks again for the transparency and the detailed cost ledger.

1 Like

I agree, it is very hard to compare results between runs since it is taking very long.

1 Like