Data Hackathon · Submission
Submit your entry
Attach your notebook (.ipynb), trained model (.joblib), requirements.txt, test features (test_x.csv), predictions, and metrics — preview them here, then submit for judging.
Submission format
The judging API actually runs your model against a hidden test set — get these right and it scores cleanly instead of coming back failed_run.
Model — .joblib only
Save it with the snippet below. It has to unpickle with the exact library versions listed in your requirements.txt — a version mismatch (e.g. training with one scikit-learn version, listing another) is the most common cause of a failed run.
joblib.dump(model, "model.joblib")requirements.txt — only what you import, pinned
List only the packages your notebook actually imports — check with the command below — pinned to the exact version you trained with. Unrelated packages just slow down environment setup on the judge's side.
pip freeze | grep -iE "scikit-learn|numpy|pandas|joblib"predictions.csv — your model's predictions on the test set
metrics.csv — your own self-reported evaluation
Problem statement & Submission guideline
Read this before you start — it explains the columns in the test CSV and the prediction format expected.
Loading…
Submission details
Choose what you're submitting, then attach your notebook and predictions below.
Pinned dependencies the judging API builds your environment from before running your model.
The test features your model runs on — sent to the judging API with your submission.
Your own self-reported evaluation of the model above — accuracy, precision, recall, f1, and anything else relevant. A sanity check for the admin, not what determines your score.