Runtime Restriction Questions


Just a few final questions about the runtime.

  1. Do we need to include our explainability solution as part of this submission? Or is this something we do later?

  2. Can we use pre-trained models or pre-uploaded data in our submission? This would substantially cut down on time needed to run the code when it is executed during the forecast period.

Hi @jitters,

I’m sure they’ll have no problems correcting me if my information is wrong, but I’m pretty sure I can answer these for you:

  1. No, that will be part of later stages. The details of how are still being discussed, I believe.
  2. Absolutely! There’s been no mention of any limitations, but I’m sure if your model weights are several hundred megabytes, you may run into problems. With the exception of processing up-to-the-minute data by ingesting it, you are both allowed and expected to be including trained model files along with your code submission.

Best of luck,

1 Like

Hi @jitters, @mmiron,

  1. Explainability submission: @mmiron is correct. Submissions will be accepted for this in a later stage of the challenge. There will be a separate submission acceptor and explicit instructions for this.

  2. Training should happen outside of the code execution. You are expected to upload trained model weights and fitted feature parameters based on the training set. There is no training data provided in the mounted data directory—only data associated with the test set water years. However, data processing needed for inference on the test set is expected to happen in the code execution runtime. Please see this recently added FAQ entry for additional discussion.

@jitters @mmiron We have recently added a preview of later stages of the challenge, including some information about the Explainability bonus track. Please see the latest announcement.