Chronos STL Attention Invariance — Raw Benchmark Datasets for Reproducibility

Raw benchmark datasets archived to ensure stable, unauthenticated access for reproducibility. No modifications have been made to any file. These are the original raw files from their respective canonical sources: PJME_hourly.csv Source: PJM Interconnection LLC, via Kaggle dataset "Hourly Energy Consumption" by Rob Mulla (https://www.kaggle.com/datasets/robikscube/hourly-energy-consumption) IEEE DataPort DOI: 10.21227/wqg6-9917 Content: Hourly electricity demand for the PJM East region, 2002–2018. ETTm1.csv Source: Zhou et al., "Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting", AAAI 2021 (Best Paper Award). Canonical repo: https://github.com/zhouhaoyi/ETDataset Content: 15-minute electricity transformer temperature readings, July 2016 – June 2018. LD2011_2014.txt. Source: UCI Machine Learning Repository, Dataset #321, "ElectricityLoadDiagrams20112014" https://archive.ics.uci.edu/dataset/321/electricityloaddiagrams20112014 Contributor: Trindade, A. (Energias de Portugal / EDP) Citation: Dua, D. and Graff, C. (2019). UCI Machine Learning Repository. Content: 15-minute electricity consumption for 370 Portuguese clients, 2011–2014.