https://athletix.run/challenges/czaMEOxQG 1st place's source code

1. Unzip dataset into /store

2. Train models from fold 0 to 3. The output score shoud be around 0.5.

# app train <fold_index> <lr> <check_interval>
app train 0 1e-3 10

3. Check training score with all models.

# app pre-submit <fold_indices>
app pre-submit 0 1 2 3

4. The target can be derived by all of these models. The inference outputs will then be saved in /store/submit.

# app submit <fold_indices>
app submit 0 1 2 3