NSF HDR ML Challenge

Codabench Tutorial

​

Yuan-Tang Chou (UW)

A3D3 Institute

for HDR ML Challenge team

​

1

PHY-2117997

https://a3d3.ai/

Reminder

• The jobs submitted must be run manually and approve by organizer team in the systems. Please inform us the time you want to run the hackathon.
• We will create a Indic for all events https://indico.cern.ch/category/19105/
• Please let me know once the date is decided. (Yuan-tang Chou ytchou@uw.edu)
• Avoid running event during NERSC Perlmutter maintenance [Link] Next one is 12/18/24! �
• Poster template: [Link]
• Google forum template [Link]
• Slack channel for local organizer support: https://hdrecosystem.slack.com/archives/C07UJUWT1D5

2

1. Login or Create Account on Codabench

3

2. Download Dummy Submission

4

3. Register in the Competition

5

4. Submit Dummy Submission

6

5. Check results in the leaderboard

7

6. Check out the starting kit

8

7. Starting kit as a Google Colab Notebook

9

8. Get Public Data

10

9. Checkout example submissions

11

11. Code submission structure

��

​

​

​

​

​

[*] Follow the example to load your model. Avoid hard-coded path to model weight

12

[Example]

Common issue

[!!] Do not zip the whole folder. ONLY select the model.py and relevant weight files to make the tarball

​

​

​

​

If you see the above error, mostly likely you zip the whole folder when making the tarball

13

14

ML Challenges