YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Experiment 4 v1 — AutoGluon MOS Bias Corrector
Overview
AutoGluon tabular model for calibrating DiT temperature predictions to Weather Underground ground truth. Maps raw DiT daily high predictions + meteorological features to calibrated decimal Celsius.
Results (OOF on 2020-2024)
| Config | OOF MAE (C) | OOF Accuracy (%) |
|---|---|---|
| resid_baseline_nobag | 0.979 | 33.8 |
| resid_compact_bag10_stack3 | 1.252 | 26.6 |
| resid_warmstart_bag10_stack3 | 1.253 | 27.9 |
Training Config
- Preset: extreme
- Bagging: 10 folds, 3 stack levels
- Time limit: 30 min per config
- Target: observed_high_c (F-precision decimal Celsius from WU)
- Training: 2020-2024 (OOD from foundation model)
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support