Submission of relevant questions

[WeiFeng Liu]

Hello, I am now ready to submit my solution, in this year's competition I am using deep learning to solve this problem, hence my model file is larger, I see in the submission guidelines that you will train our model on the test side and then validate and test it. So do we not have to submit locally trained weights?

[PhysioNet Challenge]

Dear WeiFeng,

Please submit both your training code and the model or weights from your training code, which helps us to check the full reproducibility of your code.

If you are unable to submit your trained model because the model size is too large, then please respond to us directly with the size of your model so that we can better understand the problem and try to help. (We respond to some questions directly to help protect the independence of the approaches.)

We deliberately request and especially appreciate this kind of feedback during the unofficial phase of the Challenge because it helps us to make changes to improve the official phase, so thank you for sharing it.

Best,
Matt
(On behalf of the Challenge team.)

Please post questions and comments in the forum. However, if your question reveals information about your entry, then please email info at physionetchallenge.org. We may post parts of our reply publicly if we feel that all Challengers should benefit from it. We will not answer emails about the Challenge to any other address. This email is maintained by a group. Please do not email us individually.

[xnej...@gmail.com]

Dear Matt,

Could you please clarify following: "submit both your training code and the model or weights from your training code, which helps us to check the full reproducibility of your code".

During inference on the test set, will you evaluate the submitted model (with its original submitted weights) and then check if the submitted code can resume training from those weights?

Or will you instead evaluate a "server-side" version of the model (server side optimized weights)?

Thanks,

Petr

Dne úterý 4. března 2025 v 4:01:47 UTC+1 uživatel PhysioNet Challenge napsal:

[PhysioNet Challenge]

Dear Petr,

Good question. We are currently evaluating the "server-side" version of your model, but we are also asking you to submit your "local" version of your model, which helps us to debug various problems that may arise with the submitted code and our pipeline for running it.

To more directly answer your question, we are running the sequence of commands (train_model, run_model, evaluate_model) in this section of this README using your submitted code:
https://github.com/physionetchallenges/python-example-2025?tab=readme-ov-file#how-do-i-run-these-scripts

If your submitted code trains a model, saves a model, loads the same model, and then performs inference with the model, then that is what we will do server side. If it does something else, e.g., transfer learning, then that is also what we will do, but we still require working training code that actually learns from the training set.

Best,
Matt
(On behalf of the Challenge team.)

Please post questions and comments in the forum. However, if your question reveals information about your entry, then please email info at physionetchallenge.org. We may post parts of our reply publicly if we feel that all Challengers should benefit from it. We will not answer emails about the Challenge to any other address. This email is maintained by a group. Please do not email us individually.