Solution for the Jane Street 2024 Kaggle competition.

A detailed description can be found in solution.md and in Kaggle discussion.

• ~100GB RAM
• 12GB GPU RAM

1. Install requirements from pyproject.toml.
2. Download the dataset from Kaggle.
3. Set paths and other config variables in janestreet/config.py.

• run_cv.py - Estimate model on cross-validation.
• run_full.py - Estimate model for the final submission (on the whole sample).
• run_ensemble.py - Evaluate ensemble of models on CV.
• run_test_gap.py - Test model on a sample of the last 200 dates with a gap of 200 dates.

• monitor_kaggle.py - Monitor kaggle submissions and send notifications when completed.
• update_kaggle.py - Push code and models to Kaggle datasets to be used in submission.