Hirondel 1.0.0

.gitattributes

@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+showcase/figures/banner.png filter=lfs diff=lfs merge=lfs -text
+showcase/figures/indeval.png filter=lfs diff=lfs merge=lfs -text
+showcase/figures/pareto.png filter=lfs diff=lfs merge=lfs -text
+showcase/figures/same_protocol.png filter=lfs diff=lfs merge=lfs -text
+showcase/figures/univariate.png filter=lfs diff=lfs merge=lfs -text

DATA.md

@@ -0,0 +1,572 @@
+# Training data
+
+The model was pretrained on 56 real-data sources, listed below with their holder, license, origin and the attribution each license asks for, and on synthetic series.
+
+## Synthetic data
+
+About 50% of the training windows were synthetic series. How they were generated is not described in this release.
+
+## Real data
+
+| Source | Domain | Holder | License |
+|---|---|---|---|
+| 6-second load measurement dataset (Borealis) | energy | Omid Ardakanian, S. Huo, S. Keshav | CC0-1.0 |
+| AEMO regional electricity demand (VIC1, 2014) | energy | Australian Energy Market Operator (AEMO) | AEMO general copyright permission (use for any purpose, with attribution) |
+| AEMO wind farms (NEMweb) | energy | Australian Energy Market Operator (AEMO) | AEMO general copyright permission (use for any purpose, with attribution) |
+| AEMO wind power (NEMweb) | energy | Australian Energy Market Operator (AEMO) | AEMO general copyright permission (use for any purpose, with attribution) |
+| Azure Functions traces (2019, 2021) | cloud computing | Microsoft | CC-BY-4.0 |
+| Azure Public Dataset V1 (VM traces, 2017) | cloud computing | Microsoft | CC-BY-4.0 |
+| Bosch CNC machining data | industry | Mohamed-Ali Tnani, Michael Feil, Klaus Diepold | CC-BY-4.0 |
+| Building Data Genome Project 2 (Bear site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Building Data Genome Project 2 (Bull site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Building Data Genome Project 2 (Cockatoo site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Building Data Genome Project 2 (Fox site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Building Data Genome Project 2 (Hog site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Building Data Genome Project 2 (Panther site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Building Data Genome Project 2 (Rat site) | energy | Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers | CC-BY-4.0 |
+| Buildings-900K (BuildingsBench) | energy | National Renewable Energy Laboratory (NREL); creators: Patrick Emami, Peter Graf | CC-BY-4.0 |
+| Caltrans PeMS (hourly and weekly traffic) | road traffic | State of California | Public domain (Caltrans PeMS Use Policy) |
+| Caltrans PeMS (PEMS-BAY subset) | road traffic | State of California | Public domain (Caltrans PeMS Use Policy) |
+| Caltrans PeMS (PEMS03 subset) | road traffic | State of California | Public domain (Caltrans PeMS Use Policy) |
+| Caltrans PeMS (PEMS04 subset) | road traffic | State of California | Public domain (Caltrans PeMS Use Policy) |
+| Caltrans PeMS (PEMS07 subset) | road traffic | State of California | Public domain (Caltrans PeMS Use Policy) |
+| Caltrans PeMS (PEMS08 subset) | road traffic | State of California | Public domain (Caltrans PeMS Use Policy) |
+| CDC FluView ILINet | health | U.S. Centers for Disease Control and Prevention (CDC) | Public domain (U.S. federal government work, 17 U.S.C. §105) |
+| CMIP6, MPI-ESM1.2-HR (historical) | climate | Max Planck Institute for Meteorology (MPI-M); creators: J. Jungclaus et al. (48 creators) | CC-BY-4.0 |
+| Condition monitoring of hydraulic systems (ZeMA, UCI) | industry | Nikolai Helwig, Eliseo Pignanelli, Andreas Schütze | CC-BY-4.0 |
+| EIA-930 Hourly Electric Grid Monitor | energy | U.S. Energy Information Administration (EIA) | Public domain: "U.S. government publications are in the public domain and are not subject to copyright protection." |
+| EPA Air Quality System (AQS) | air quality | U.S. Environmental Protection Agency (EPA) | Public domain: "The ambient monitoring data in EPA's Air Quality System (AQS) are public domain." |
+| ERA5 reanalysis | climate | European Centre for Medium-Range Weather Forecasts (ECMWF) | CC-BY-4.0 |
+| Eurostat short-term business statistics | economics | European Union | Eurostat re-use policy: re-use "for commercial or non-commercial purposes is authorised provided the source is acknowledged" (European Commission decision of 12 December 2011) |
+| Google Borg cluster trace (2011) | cloud computing | Google | CC-BY-4.0 |
+| Google PowerData2019 | cloud computing | Google | CC-BY-4.0 |
+| IDEAL Household Energy Dataset | energy | University of Edinburgh, School of Informatics (publisher); creators: N. Goddard et al. (18 creators) | CC-BY-4.0 |
+| Individual household electric power consumption (UCI) | energy | Georges Hebrail, Alice Berard | CC-BY-4.0 |
+| London smart meter data (per household) | energy | UK Power Networks | CC-BY-4.0 |
+| Low Carbon London smart meter data (LCL) | energy | UK Power Networks | CC-BY-4.0 |
+| M1 competition data | economics | International Institute of Forecasters (IIF) | IIF general permission: "freely available for anyone to use" (not a named licence) |
+| M3 competition data | economics | International Institute of Forecasters (IIF) | IIF general permission: "freely available for anyone to use" (not a named licence) |
+| M3 competition data (other) | economics | International Institute of Forecasters (IIF) | IIF general permission: "freely available for anyone to use" (not a named licence) |
+| Melbourne Pedestrian Counting System | mobility | City of Melbourne | CC-BY-4.0 |
+| N2O in wastewater treatment (Aalborg University) | industry | Laura Debel Hansen, Anju Rani, Daniel Ortiz Arroyo, Petar Durdevic | CC-BY-4.0 |
+| National Highways WebTRIS traffic counts | road traffic | National Highways (Crown copyright) | OGL-UK-3.0 (Open Government Licence v3.0) |
+| NOAA CO-OPS hourly water levels | ocean | NOAA National Ocean Service, Center for Operational Oceanographic Products and Services (CO-OPS) | Public domain: "The information on government servers are in the public domain, unless specifically annotated otherwise, and may be used freely by the public." |
+| NREL PVDAQ public datasets | energy | National Renewable Energy Laboratory (NREL); creators: C. Deline et al. | CC-BY-4.0 |
+| Penmanshiel and Kelmarsh wind farm data (Cubico) | industry | Cubico Sustainable Investments Ltd; creators: Charlie Plumley, Roberta Takeuchi | CC-BY-4.0 |
+| Project Tycho | health | University of Pittsburgh | CC-BY-4.0 |
+| Residential Power and Battery Data (load) | energy | Christoph Bergmeir, Quang Bui, Frits de Nijs, Peter Stuckey (Monash University); sponsor: SwitchDin Ltd | CC-BY-4.0 |
+| Residential Power and Battery Data (PV) | energy | Christoph Bergmeir, Quang Bui, Frits de Nijs, Peter Stuckey (Monash University); sponsor: SwitchDin Ltd | CC-BY-4.0 |
+| Statistics Canada tables | economics | His Majesty the King in Right of Canada, as represented by the Minister for Statistics Canada | Statistics Canada Open Licence |
+| Steel Industry Energy Consumption (UCI) | industry | V E Sathishkumar, Changsun Shin, Yongyun Cho | CC-BY-4.0 |
+| SubseasonalClimateUSA | climate | Soukayna Mouatadid, Paulo Orenstein, Genevieve Flaspohler, Miruna Oprescu, Judah Cohen, Franklyn Wang, Sean Knight, Maria Geogdzhayeva, Sam Levang, Ernest Fraenkel, Lester Mackey | CC-BY-4.0 |
+| SubseasonalClimateUSA (precipitation) | climate | Soukayna Mouatadid, Paulo Orenstein, Genevieve Flaspohler, Miruna Oprescu, Judah Cohen, Franklyn Wang, Sean Knight, Maria Geogdzhayeva, Sam Levang, Ernest Fraenkel, Lester Mackey | CC-BY-4.0 |
+| Transport for NSW permanent hourly traffic counts | road traffic | Transport for NSW | CC-BY-4.0 |
+| UMass Smart* data set | energy | University of Massachusetts Amherst | CC-BY-4.0 |
+| USGS NWIS instantaneous streamflow | hydrology | U.S. Geological Survey (USGS) | Public domain: "USGS-authored or produced data and information are considered to be in the U.S. Public Domain." |
+| Wikipedia page views (extended web traffic) | web | not required (CC0) | CC0-1.0 |
+| Wikipedia page views (rolling) | web | not required (CC0) | CC0-1.0 |
+| Wikipedia page views (weekly web traffic) | web | not required (CC0) | CC0-1.0 |
+
+### 6-second load measurement dataset (Borealis)
+
+- Holder: Omid Ardakanian, S. Huo, S. Keshav
+- License: CC0-1.0
+- Origin: https://borealisdata.ca/dataset.xhtml?persistentId=doi:10.5683/SP2/R4SVBF
+- DOI: 10.5683/SP2/R4SVBF
+- Attribution: None required (CC0)
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### AEMO regional electricity demand (VIC1, 2014)
+
+- Holder: Australian Energy Market Operator (AEMO)
+- License: AEMO general copyright permission (use for any purpose, with attribution)
+- Origin: https://aemo.com.au/privacy-and-legal-notices/copyright-permissions
+- DOI: none
+- Attribution: AEMO (Australian Energy Market Operator), regional demand VIC1 2014
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### AEMO wind farms (NEMweb)
+
+- Holder: Australian Energy Market Operator (AEMO)
+- License: AEMO general copyright permission (use for any purpose, with attribution)
+- Origin: https://aemo.com.au/privacy-and-legal-notices/copyright-permissions
+- DOI: none
+- Attribution: AEMO (Australian Energy Market Operator), NEMweb
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### AEMO wind power (NEMweb)
+
+- Holder: Australian Energy Market Operator (AEMO)
+- License: AEMO general copyright permission (use for any purpose, with attribution)
+- Origin: https://aemo.com.au/privacy-and-legal-notices/copyright-permissions
+- DOI: none
+- Attribution: AEMO (Australian Energy Market Operator), NEMweb
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Azure Functions traces (2019, 2021)
+
+- Holder: Microsoft
+- License: CC-BY-4.0
+- Origin: https://github.com/Azure/AzurePublicDataset/blob/master/AzureFunctionsDataset2019.md
+- DOI: none
+- Attribution: Microsoft Azure, Azure Public Dataset: Azure Functions Trace 2019 (M. Shahrad et al., "Serverless in the Wild", USENIX ATC 2020) and Azure Functions Invocation Trace 2021 (Y. Zhang et al., "Faster and Cheaper Serverless Computing on Harvested Resources", SOSP 2021), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Invocations per minute of each function (2021 trace: counted by end minute); a missing day is treated as missing; 4,000 of about 72,000 functions of the 2019 trace were sampled.
+
+### Azure Public Dataset V1 (VM traces, 2017)
+
+- Holder: Microsoft
+- License: CC-BY-4.0
+- Origin: https://github.com/Azure/AzurePublicDataset
+- DOI: none
+- Attribution: Microsoft Azure, Azure Public Dataset V1 (SOSP'17), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Bosch CNC machining data
+
+- Holder: Mohamed-Ali Tnani, Michael Feil, Klaus Diepold
+- License: CC-BY-4.0
+- Origin: https://github.com/boschresearch/CNC_Machining
+- DOI: none
+- Attribution: M.-A. Tnani, M. Feil, K. Diepold, Smart Data Collection System for Brownfield CNC Milling Machines: A New Benchmark Dataset for Data-Driven Machine Monitoring, Procedia CIRP 107 (2022) 131-136, doi:10.1016/j.procir.2022.04.022; data CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Each channel was placed on a regular time grid. The 2 kHz vibration signal was reduced to an RMS envelope (50 ms); runs of the same operation were concatenated.
+
+### Building Data Genome Project 2 (Bear site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Building Data Genome Project 2 (Bull site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Building Data Genome Project 2 (Cockatoo site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Building Data Genome Project 2 (Fox site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Building Data Genome Project 2 (Hog site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Building Data Genome Project 2 (Panther site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Building Data Genome Project 2 (Rat site)
+
+- Holder: Clayton Miller, Anjukan Kathirgamanathan, Bianca Picchetti, Pandarasamy Arjunan, June Young Park, Zoltan Nagy, Paul Raftery, Brodie W. Hobson, Zixiao Shi, Forrest Meggers
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/3887306
+- DOI: 10.5281/zenodo.3887306
+- Attribution: C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Buildings-900K (BuildingsBench)
+
+- Holder: National Renewable Energy Laboratory (NREL); creators: Patrick Emami, Peter Graf
+- License: CC-BY-4.0
+- Origin: https://data.openei.org/submissions/5859
+- DOI: 10.25984/1986147
+- Attribution: P. Emami, P. Graf, BuildingsBench: A Large-Scale Dataset of 900K Buildings and Benchmark for Short-Term Load Forecasting, National Renewable Energy Laboratory, Open Energy Data Initiative (OEDI), doi:10.25984/1986147, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Caltrans PeMS (hourly and weekly traffic)
+
+- Holder: State of California
+- License: Public domain (Caltrans PeMS Use Policy)
+- Origin: https://pems.dot.ca.gov/?view=tou
+- DOI: none
+- Attribution: None required (public domain); credit: Caltrans PeMS
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Caltrans PeMS (PEMS-BAY subset)
+
+- Holder: State of California
+- License: Public domain (Caltrans PeMS Use Policy)
+- Origin: https://pems.dot.ca.gov/?view=tou
+- DOI: none
+- Attribution: None required (public domain); credit: Caltrans PeMS
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Caltrans PeMS (PEMS03 subset)
+
+- Holder: State of California
+- License: Public domain (Caltrans PeMS Use Policy)
+- Origin: https://pems.dot.ca.gov/?view=tou
+- DOI: none
+- Attribution: None required (public domain); credit: Caltrans PeMS
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Caltrans PeMS (PEMS04 subset)
+
+- Holder: State of California
+- License: Public domain (Caltrans PeMS Use Policy)
+- Origin: https://pems.dot.ca.gov/?view=tou
+- DOI: none
+- Attribution: None required (public domain); credit: Caltrans PeMS
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Caltrans PeMS (PEMS07 subset)
+
+- Holder: State of California
+- License: Public domain (Caltrans PeMS Use Policy)
+- Origin: https://pems.dot.ca.gov/?view=tou
+- DOI: none
+- Attribution: None required (public domain); credit: Caltrans PeMS
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Caltrans PeMS (PEMS08 subset)
+
+- Holder: State of California
+- License: Public domain (Caltrans PeMS Use Policy)
+- Origin: https://pems.dot.ca.gov/?view=tou
+- DOI: none
+- Attribution: None required (public domain); credit: Caltrans PeMS
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### CDC FluView ILINet
+
+- Holder: U.S. Centers for Disease Control and Prevention (CDC)
+- License: Public domain (U.S. federal government work, 17 U.S.C. §105)
+- Origin: https://gis.cdc.gov/grasp/fluview/fluportaldashboard.html
+- DOI: none
+- Attribution: None required (public domain); credit: CDC
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### CMIP6, MPI-ESM1.2-HR (historical)
+
+- Holder: Max Planck Institute for Meteorology (MPI-M); creators: J. Jungclaus et al. (48 creators)
+- License: CC-BY-4.0
+- Origin: https://pcmdi.llnl.gov/CMIP6/TermsOfUse/TermsOfUse6-2.html
+- DOI: 10.22033/ESGF/CMIP6.6594
+- Attribution: J. Jungclaus et al. (2019), MPI-M MPI-ESM1.2-HR model output prepared for CMIP6 CMIP historical, Earth System Grid Federation, doi:10.22033/ESGF/CMIP6.6594, CC BY 4.0. We acknowledge the World Climate Research Programme, which, through its Working Group on Coupled Modelling, coordinated and promoted CMIP6. We thank the climate modeling groups for producing and making available their model output, the Earth System Grid Federation (ESGF) for archiving the data and providing access, and the multiple funding agencies who support CMIP6 and ESGF.
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Condition monitoring of hydraulic systems (ZeMA, UCI)
+
+- Holder: Nikolai Helwig, Eliseo Pignanelli, Andreas Schütze
+- License: CC-BY-4.0
+- Origin: https://archive.ics.uci.edu/dataset/447/condition+monitoring+of+hydraulic+systems
+- DOI: 10.24432/C5CW21
+- Attribution: N. Helwig, E. Pignanelli, A. Schütze (ZeMA gGmbH), Condition monitoring of hydraulic systems, UCI Machine Learning Repository, doi:10.24432/C5CW21, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Each channel was placed on a regular time grid. 60-second cycles were concatenated in row order; 100 Hz and 10 Hz channels were reduced to 1 Hz (mean per second).
+
+### EIA-930 Hourly Electric Grid Monitor
+
+- Holder: U.S. Energy Information Administration (EIA)
+- License: Public domain: "U.S. government publications are in the public domain and are not subject to copyright protection."
+- Origin: https://www.eia.gov/about/copyrights_reuse.php
+- DOI: none
+- Attribution: None required (public domain); acknowledgment requested ("you should use an acknowledgment"): Source: U.S. Energy Information Administration, Hourly Electric Grid Monitor (EIA-930), Oct 2026
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Adjusted demand per balancing authority, 2015-07 to 2025-09-30; the ERCO balancing authority was excluded.
+
+### EPA Air Quality System (AQS)
+
+- Holder: U.S. Environmental Protection Agency (EPA)
+- License: Public domain: "The ambient monitoring data in EPA's Air Quality System (AQS) are public domain."
+- Origin: https://www.epa.gov/outdoor-air-quality-data/do-i-need-request-permission-use-monitoring-data-and-graphics-airdata
+- DOI: none
+- Attribution: None required (public domain); citation requested by the EPA: US Environmental Protection Agency. Air Quality System Data Mart [internet database] available via https://www.epa.gov/outdoor-air-quality-data. Accessed October 19, 2026.
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. PM2.5, ozone and NO2, 2019 to 2025-09-30; about one monitor in eight was sampled.
+
+### ERA5 reanalysis
+
+- Holder: European Centre for Medium-Range Weather Forecasts (ECMWF)
+- License: CC-BY-4.0
+- Origin: https://confluence.ecmwf.int/x/_hpbIg
+- DOI: none
+- Attribution: Modified from ECMWF Reanalysis data (ERA5), © European Centre for Medium-Range Weather Forecasts (ECMWF), 2026, accessed via the GiftEvalPretrain dataset (Salesforce, Hugging Face). Licensed under CC-BY-4.0.
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Data before 2025-10 only.
+
+### Eurostat short-term business statistics
+
+- Holder: European Union
+- License: Eurostat re-use policy: re-use "for commercial or non-commercial purposes is authorised provided the source is acknowledged" (European Commission decision of 12 December 2011)
+- Origin: https://ec.europa.eu/eurostat/web/main/help/copyright-notice
+- DOI: none
+- Attribution: Source: Eurostat, sts_inpr_m, sts_trtu_m, sts_copr_m (accessed 2026-10-20); modified data (raw monthly series extracted, split at missing values); Eurostat is not responsible for this re-use
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Only data of EU, EFTA and candidate countries (and EU / euro area aggregates) were used.
+
+### Google Borg cluster trace (2011)
+
+- Holder: Google
+- License: CC-BY-4.0
+- Origin: https://github.com/google/cluster-data
+- DOI: none
+- Attribution: Google, cluster-data 2011, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Google PowerData2019
+
+- Holder: Google
+- License: CC-BY-4.0
+- Origin: https://github.com/google/cluster-data/blob/master/PowerData2019.md
+- DOI: none
+- Attribution: Google, cluster-data PowerData2019 (V. Sakalkar et al., "Data Center Power Oversubscription with a Medium Voltage Power Plane and Priority-Aware Capping", ASPLOS 2020), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Measured and production utilisation of each power distribution unit, on a 5-minute grid; measurements flagged as dubious were removed.
+
+### IDEAL Household Energy Dataset
+
+- Holder: University of Edinburgh, School of Informatics (publisher); creators: N. Goddard et al. (18 creators)
+- License: CC-BY-4.0
+- Origin: https://datashare.ed.ac.uk/handle/10283/3647
+- DOI: 10.7488/ds/2836
+- Attribution: University of Edinburgh DataShare, IDEAL, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Individual household electric power consumption (UCI)
+
+- Holder: Georges Hebrail, Alice Berard
+- License: CC-BY-4.0
+- Origin: https://archive.ics.uci.edu/dataset/235/individual+household+electric+power+consumption
+- DOI: 10.24432/C58K54
+- Attribution: UCI Machine Learning Repository, Individual household electric power consumption, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### London smart meter data (per household)
+
+- Holder: UK Power Networks
+- License: CC-BY-4.0
+- Origin: https://data.london.gov.uk/dataset/smartmeter-energy-use-data-in-london-households
+- DOI: none
+- Attribution: UK Power Networks, London Datastore, SmartMeter Energy Consumption Data in London Households, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. One household is one series.
+
+### Low Carbon London smart meter data (LCL)
+
+- Holder: UK Power Networks
+- License: CC-BY-4.0
+- Origin: https://data.london.gov.uk/dataset/smartmeter-energy-use-data-in-london-households
+- DOI: none
+- Attribution: UK Power Networks, London Datastore, SmartMeter Energy Use Data in London Households, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### M1 competition data
+
+- Holder: International Institute of Forecasters (IIF)
+- License: IIF general permission: "freely available for anyone to use" (not a named licence)
+- Origin: https://web.archive.org/web/20260504053121id_/https://forecasters.org/resources/time-series-data/
+- DOI: none
+- Attribution: International Institute of Forecasters (IIF), M competition data
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### M3 competition data
+
+- Holder: International Institute of Forecasters (IIF)
+- License: IIF general permission: "freely available for anyone to use" (not a named licence)
+- Origin: https://web.archive.org/web/20260504053121id_/https://forecasters.org/resources/time-series-data/
+- DOI: none
+- Attribution: International Institute of Forecasters (IIF), M competition data
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### M3 competition data (other)
+
+- Holder: International Institute of Forecasters (IIF)
+- License: IIF general permission: "freely available for anyone to use" (not a named licence)
+- Origin: https://web.archive.org/web/20260504053121id_/https://forecasters.org/resources/time-series-data/
+- DOI: none
+- Attribution: International Institute of Forecasters (IIF), M competition data
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Melbourne Pedestrian Counting System
+
+- Holder: City of Melbourne
+- License: CC-BY-4.0
+- Origin: https://data.melbourne.vic.gov.au/pages/about-our-data/ ; https://web.archive.org/web/20200418070248id_/https://data.melbourne.vic.gov.au/Transport/Pedestrian-Counting-System-2009-to-Present-counts-/b2ak-trbp
+- DOI: none
+- Attribution: City of Melbourne, Pedestrian Counting System, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Local hours were converted to UTC.
+
+### N2O in wastewater treatment (Aalborg University)
+
+- Holder: Laura Debel Hansen, Anju Rani, Daniel Ortiz Arroyo, Petar Durdevic
+- License: CC-BY-4.0
+- Origin: https://data.mendeley.com/datasets/xmbxhscgpr/4
+- DOI: 10.17632/xmbxhscgpr.4
+- Attribution: L. D. Hansen, A. Rani, D. Ortiz Arroyo, P. Durdevic (Aalborg University), Time Series Dataset for Modeling and Forecasting of N2O in Wastewater Treatment, Mendeley Data v4, doi:10.17632/xmbxhscgpr.4, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Each channel was placed on a regular time grid. Values with a non-zero quality flag were removed.
+
+### National Highways WebTRIS traffic counts
+
+- Holder: National Highways (Crown copyright)
+- License: OGL-UK-3.0 (Open Government Licence v3.0)
+- Origin: https://www.data.gov.uk/dataset/9562c512-4a0b-45ee-b6ad-afc0f99b841f/highways-england-network-journey-time-and-traffic-flow-data
+- DOI: none
+- Attribution: Contains public sector information licensed under the Open Government Licence v3.0 (National Highways, WebTRIS, TMU / TAME sites, 15-minute counts)
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Data up to 2025-09-30 only.
+
+### NOAA CO-OPS hourly water levels
+
+- Holder: NOAA National Ocean Service, Center for Operational Oceanographic Products and Services (CO-OPS)
+- License: Public domain: "The information on government servers are in the public domain, unless specifically annotated otherwise, and may be used freely by the public."
+- Origin: https://tidesandcurrents.noaa.gov/disclaimers.html
+- DOI: none
+- Attribution: None required (public domain); attribution requested ("NOS requests that attribution be given whenever NOS material is reproduced and re-disseminated"): NOAA National Ocean Service, CO-OPS, verified hourly water levels (Tides & Currents)
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Verified hourly heights, 2010-10 to 2025-09-30; 20 stations were excluded.
+
+### NREL PVDAQ public datasets
+
+- Holder: National Renewable Energy Laboratory (NREL); creators: C. Deline et al.
+- License: CC-BY-4.0
+- Origin: https://data.openei.org/submissions/4568
+- DOI: 10.25984/1846021
+- Attribution: C. Deline, K. Perry, M. Deceglie, M. Muller, W. Sekulic, D. Jordan (NREL), Photovoltaic Data Acquisition (PVDAQ) Public Datasets, Open Energy Data Initiative (OEDI), doi:10.25984/1846021, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Each channel was placed on a regular time grid. Systems 2105, 2107 and 9068 of the 2023 DOE Solar Data Prize only; timestamps without a time zone were taken as UTC; nights without measurements were filled with 0 on production channels (added values).
+
+### Penmanshiel and Kelmarsh wind farm data (Cubico)
+
+- Holder: Cubico Sustainable Investments Ltd; creators: Charlie Plumley, Roberta Takeuchi
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/16807304 ; https://zenodo.org/records/16807551
+- DOI: 10.5281/zenodo.16807304 ; 10.5281/zenodo.16807551
+- Attribution: Cubico Sustainable Investments Ltd; C. Plumley, R. Takeuchi, Penmanshiel wind farm data (Zenodo, doi:10.5281/zenodo.16807304) and Kelmarsh wind farm data (Zenodo, doi:10.5281/zenodo.16807551), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Each channel was placed on a regular time grid. Sample: Penmanshiel turbines WT11-15 and Kelmarsh turbines WT1-6, 2016-2019; mean physical channels only.
+
+### Project Tycho
+
+- Holder: University of Pittsburgh
+- License: CC-BY-4.0
+- Origin: https://www.tycho.pitt.edu/license_agreement_text
+- DOI: none
+- Attribution: Project Tycho, University of Pittsburgh, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Residential Power and Battery Data (load)
+
+- Holder: Christoph Bergmeir, Quang Bui, Frits de Nijs, Peter Stuckey (Monash University); sponsor: SwitchDin Ltd
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/8219786
+- DOI: 10.5281/zenodo.8219786
+- Attribution: SwitchDin / Monash University, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Residential Power and Battery Data (PV)
+
+- Holder: Christoph Bergmeir, Quang Bui, Frits de Nijs, Peter Stuckey (Monash University); sponsor: SwitchDin Ltd
+- License: CC-BY-4.0
+- Origin: https://zenodo.org/records/8219786
+- DOI: 10.5281/zenodo.8219786
+- Attribution: SwitchDin / Monash University, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Statistics Canada tables
+
+- Holder: His Majesty the King in Right of Canada, as represented by the Minister for Statistics Canada
+- License: Statistics Canada Open Licence
+- Origin: https://www.statcan.gc.ca/en/reference/licence
+- DOI: none
+- Attribution: Adapted from Statistics Canada, tables 16-10-0047-01, 20-10-0008-01, 25-10-0015-01 (2026-10-20). This does not constitute an endorsement by Statistics Canada of this product.
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Steel Industry Energy Consumption (UCI)
+
+- Holder: V E Sathishkumar, Changsun Shin, Yongyun Cho
+- License: CC-BY-4.0
+- Origin: https://archive.ics.uci.edu/dataset/851/steel+industry+energy+consumption
+- DOI: 10.24432/C52G8C
+- Attribution: V E Sathishkumar, C. Shin, Y. Cho, Steel Industry Energy Consumption, UCI Machine Learning Repository, doi:10.24432/C52G8C, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Each channel was placed on a regular time grid.
+
+### SubseasonalClimateUSA
+
+- Holder: Soukayna Mouatadid, Paulo Orenstein, Genevieve Flaspohler, Miruna Oprescu, Judah Cohen, Franklyn Wang, Sean Knight, Maria Geogdzhayeva, Sam Levang, Ernest Fraenkel, Lester Mackey
+- License: CC-BY-4.0
+- Origin: https://github.com/microsoft/subseasonal_data
+- DOI: none
+- Attribution: S. Mouatadid et al., SubseasonalClimateUSA: A Dataset for Subseasonal Forecasting and Benchmarking, Advances in Neural Information Processing Systems 36 (NeurIPS 2023), CC BY 4.0; the subseasonal_data package (https://github.com/microsoft/subseasonal_data); and the individual data sources listed in https://github.com/microsoft/subseasonal_data/blob/main/DATA.md (Data Sources and Citation), including NOAA/OAR/ESRL PSL CPC global temperature data and CPC global and US unified precipitation data
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### SubseasonalClimateUSA (precipitation)
+
+- Holder: Soukayna Mouatadid, Paulo Orenstein, Genevieve Flaspohler, Miruna Oprescu, Judah Cohen, Franklyn Wang, Sean Knight, Maria Geogdzhayeva, Sam Levang, Ernest Fraenkel, Lester Mackey
+- License: CC-BY-4.0
+- Origin: https://github.com/microsoft/subseasonal_data
+- DOI: none
+- Attribution: S. Mouatadid et al., SubseasonalClimateUSA: A Dataset for Subseasonal Forecasting and Benchmarking, Advances in Neural Information Processing Systems 36 (NeurIPS 2023), CC BY 4.0; the subseasonal_data package (https://github.com/microsoft/subseasonal_data); and the individual data sources listed in https://github.com/microsoft/subseasonal_data/blob/main/DATA.md (Data Sources and Citation), including NOAA/OAR/ESRL PSL CPC global temperature data and CPC global and US unified precipitation data
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Transport for NSW permanent hourly traffic counts
+
+- Holder: Transport for NSW
+- License: CC-BY-4.0
+- Origin: https://opendata.transport.nsw.gov.au/datalicence
+- DOI: none
+- Attribution: Transport for NSW, NSW Roads Traffic Volume Counts (Permanent Hourly Traffic Counts), CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Data up to 2025-09-30 only.
+
+### UMass Smart* data set
+
+- Holder: University of Massachusetts Amherst
+- License: CC-BY-4.0
+- Origin: https://traces.cs.umass.edu/docs/traces/smartstar/
+- DOI: none
+- Attribution: UMass Trace Repository, Smart*, CC BY 4.0
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### USGS NWIS instantaneous streamflow
+
+- Holder: U.S. Geological Survey (USGS)
+- License: Public domain: "USGS-authored or produced data and information are considered to be in the U.S. Public Domain."
+- Origin: https://www.usgs.gov/information-policies-and-instructions/copyrights-and-credits
+- DOI: none
+- Attribution: None required (public domain); credit requested ("we ask that proper credit be given"): U.S. Geological Survey, National Water Information System (NWIS), instantaneous streamflow values
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used. Data up to 2025-09-30 only. 30 stations were excluded.
+
+### Wikipedia page views (extended web traffic)
+
+- Holder: not required (CC0)
+- License: CC0-1.0
+- Origin: https://dumps.wikimedia.org/other/pageviews/readme.html
+- DOI: none
+- Attribution: None required (CC0); credit: Wikimedia
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Wikipedia page views (rolling)
+
+- Holder: not required (CC0)
+- License: CC0-1.0
+- Origin: https://dumps.wikimedia.org/other/pageviews/readme.html
+- DOI: none
+- Attribution: None required (CC0); credit: Wikimedia
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.
+
+### Wikipedia page views (weekly web traffic)
+
+- Holder: not required (CC0)
+- License: CC0-1.0
+- Origin: https://dumps.wikimedia.org/other/pageviews/readme.html
+- DOI: none
+- Attribution: None required (CC0); credit: Wikimedia
+- Changes: Each variable was taken as a separate univariate series and split at missing values into finite segments; a subset of the series was used.

LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

NOTICE

@@ -0,0 +1,62 @@
+Hirondel
+Copyright 2026 The Hirondel Author
+
+The model weights were trained on third-party data. The notices below are required by the licenses of
+that data; DATA.md gives the license, origin and changes for each source.
+
+- 6-second load measurement dataset (Borealis): None required (CC0)
+- AEMO regional electricity demand (VIC1, 2014): AEMO (Australian Energy Market Operator), regional demand VIC1 2014
+- AEMO wind farms (NEMweb): AEMO (Australian Energy Market Operator), NEMweb
+- AEMO wind power (NEMweb): AEMO (Australian Energy Market Operator), NEMweb
+- Azure Functions traces (2019, 2021): Microsoft Azure, Azure Public Dataset: Azure Functions Trace 2019 (M. Shahrad et al., "Serverless in the Wild", USENIX ATC 2020) and Azure Functions Invocation Trace 2021 (Y. Zhang et al., "Faster and Cheaper Serverless Computing on Harvested Resources", SOSP 2021), CC BY 4.0
+- Azure Public Dataset V1 (VM traces, 2017): Microsoft Azure, Azure Public Dataset V1 (SOSP'17), CC BY 4.0
+- Bosch CNC machining data: M.-A. Tnani, M. Feil, K. Diepold, Smart Data Collection System for Brownfield CNC Milling Machines: A New Benchmark Dataset for Data-Driven Machine Monitoring, Procedia CIRP 107 (2022) 131-136, doi:10.1016/j.procir.2022.04.022; data CC BY 4.0
+- Building Data Genome Project 2 (Bear site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Building Data Genome Project 2 (Bull site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Building Data Genome Project 2 (Cockatoo site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Building Data Genome Project 2 (Fox site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Building Data Genome Project 2 (Hog site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Building Data Genome Project 2 (Panther site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Building Data Genome Project 2 (Rat site): C. Miller et al., Building Data Genome Project 2, Zenodo archive v1.0 (2020-06-10), CC BY 4.0
+- Buildings-900K (BuildingsBench): P. Emami, P. Graf, BuildingsBench: A Large-Scale Dataset of 900K Buildings and Benchmark for Short-Term Load Forecasting, National Renewable Energy Laboratory, Open Energy Data Initiative (OEDI), doi:10.25984/1986147, CC BY 4.0
+- Caltrans PeMS (hourly and weekly traffic): None required (public domain); credit: Caltrans PeMS
+- Caltrans PeMS (PEMS-BAY subset): None required (public domain); credit: Caltrans PeMS
+- Caltrans PeMS (PEMS03 subset): None required (public domain); credit: Caltrans PeMS
+- Caltrans PeMS (PEMS04 subset): None required (public domain); credit: Caltrans PeMS
+- Caltrans PeMS (PEMS07 subset): None required (public domain); credit: Caltrans PeMS
+- Caltrans PeMS (PEMS08 subset): None required (public domain); credit: Caltrans PeMS
+- CDC FluView ILINet: None required (public domain); credit: CDC
+- CMIP6, MPI-ESM1.2-HR (historical): J. Jungclaus et al. (2019), MPI-M MPI-ESM1.2-HR model output prepared for CMIP6 CMIP historical, Earth System Grid Federation, doi:10.22033/ESGF/CMIP6.6594, CC BY 4.0. We acknowledge the World Climate Research Programme, which, through its Working Group on Coupled Modelling, coordinated and promoted CMIP6. We thank the climate modeling groups for producing and making available their model output, the Earth System Grid Federation (ESGF) for archiving the data and providing access, and the multiple funding agencies who support CMIP6 and ESGF.
+- Condition monitoring of hydraulic systems (ZeMA, UCI): N. Helwig, E. Pignanelli, A. Schütze (ZeMA gGmbH), Condition monitoring of hydraulic systems, UCI Machine Learning Repository, doi:10.24432/C5CW21, CC BY 4.0
+- EIA-930 Hourly Electric Grid Monitor: None required (public domain); acknowledgment requested ("you should use an acknowledgment"): Source: U.S. Energy Information Administration, Hourly Electric Grid Monitor (EIA-930), Oct 2026
+- EPA Air Quality System (AQS): None required (public domain); citation requested by the EPA: US Environmental Protection Agency. Air Quality System Data Mart [internet database] available via https://www.epa.gov/outdoor-air-quality-data. Accessed October 19, 2026.
+- ERA5 reanalysis: Modified from ECMWF Reanalysis data (ERA5), © European Centre for Medium-Range Weather Forecasts (ECMWF), 2026, accessed via the GiftEvalPretrain dataset (Salesforce, Hugging Face). Licensed under CC-BY-4.0.
+- Eurostat short-term business statistics: Source: Eurostat, sts_inpr_m, sts_trtu_m, sts_copr_m (accessed 2026-10-20); modified data (raw monthly series extracted, split at missing values); Eurostat is not responsible for this re-use
+- Google Borg cluster trace (2011): Google, cluster-data 2011, CC BY 4.0
+- Google PowerData2019: Google, cluster-data PowerData2019 (V. Sakalkar et al., "Data Center Power Oversubscription with a Medium Voltage Power Plane and Priority-Aware Capping", ASPLOS 2020), CC BY 4.0
+- IDEAL Household Energy Dataset: University of Edinburgh DataShare, IDEAL, CC BY 4.0
+- Individual household electric power consumption (UCI): UCI Machine Learning Repository, Individual household electric power consumption, CC BY 4.0
+- London smart meter data (per household): UK Power Networks, London Datastore, SmartMeter Energy Consumption Data in London Households, CC BY 4.0
+- Low Carbon London smart meter data (LCL): UK Power Networks, London Datastore, SmartMeter Energy Use Data in London Households, CC BY 4.0
+- M1 competition data: International Institute of Forecasters (IIF), M competition data
+- M3 competition data: International Institute of Forecasters (IIF), M competition data
+- M3 competition data (other): International Institute of Forecasters (IIF), M competition data
+- Melbourne Pedestrian Counting System: City of Melbourne, Pedestrian Counting System, CC BY 4.0
+- N2O in wastewater treatment (Aalborg University): L. D. Hansen, A. Rani, D. Ortiz Arroyo, P. Durdevic (Aalborg University), Time Series Dataset for Modeling and Forecasting of N2O in Wastewater Treatment, Mendeley Data v4, doi:10.17632/xmbxhscgpr.4, CC BY 4.0
+- National Highways WebTRIS traffic counts: Contains public sector information licensed under the Open Government Licence v3.0 (National Highways, WebTRIS, TMU / TAME sites, 15-minute counts)
+- NOAA CO-OPS hourly water levels: None required (public domain); attribution requested ("NOS requests that attribution be given whenever NOS material is reproduced and re-disseminated"): NOAA National Ocean Service, CO-OPS, verified hourly water levels (Tides & Currents)
+- NREL PVDAQ public datasets: C. Deline, K. Perry, M. Deceglie, M. Muller, W. Sekulic, D. Jordan (NREL), Photovoltaic Data Acquisition (PVDAQ) Public Datasets, Open Energy Data Initiative (OEDI), doi:10.25984/1846021, CC BY 4.0
+- Penmanshiel and Kelmarsh wind farm data (Cubico): Cubico Sustainable Investments Ltd; C. Plumley, R. Takeuchi, Penmanshiel wind farm data (Zenodo, doi:10.5281/zenodo.16807304) and Kelmarsh wind farm data (Zenodo, doi:10.5281/zenodo.16807551), CC BY 4.0
+- Project Tycho: Project Tycho, University of Pittsburgh, CC BY 4.0
+- Residential Power and Battery Data (load): SwitchDin / Monash University, CC BY 4.0
+- Residential Power and Battery Data (PV): SwitchDin / Monash University, CC BY 4.0
+- Statistics Canada tables: Adapted from Statistics Canada, tables 16-10-0047-01, 20-10-0008-01, 25-10-0015-01 (2026-10-20). This does not constitute an endorsement by Statistics Canada of this product.
+- Steel Industry Energy Consumption (UCI): V E Sathishkumar, C. Shin, Y. Cho, Steel Industry Energy Consumption, UCI Machine Learning Repository, doi:10.24432/C52G8C, CC BY 4.0
+- SubseasonalClimateUSA: S. Mouatadid et al., SubseasonalClimateUSA: A Dataset for Subseasonal Forecasting and Benchmarking, Advances in Neural Information Processing Systems 36 (NeurIPS 2023), CC BY 4.0; the subseasonal_data package (https://github.com/microsoft/subseasonal_data); and the individual data sources listed in https://github.com/microsoft/subseasonal_data/blob/main/DATA.md (Data Sources and Citation), including NOAA/OAR/ESRL PSL CPC global temperature data and CPC global and US unified precipitation data
+- SubseasonalClimateUSA (precipitation): S. Mouatadid et al., SubseasonalClimateUSA: A Dataset for Subseasonal Forecasting and Benchmarking, Advances in Neural Information Processing Systems 36 (NeurIPS 2023), CC BY 4.0; the subseasonal_data package (https://github.com/microsoft/subseasonal_data); and the individual data sources listed in https://github.com/microsoft/subseasonal_data/blob/main/DATA.md (Data Sources and Citation), including NOAA/OAR/ESRL PSL CPC global temperature data and CPC global and US unified precipitation data
+- Transport for NSW permanent hourly traffic counts: Transport for NSW, NSW Roads Traffic Volume Counts (Permanent Hourly Traffic Counts), CC BY 4.0
+- UMass Smart* data set: UMass Trace Repository, Smart*, CC BY 4.0
+- USGS NWIS instantaneous streamflow: None required (public domain); credit requested ("we ask that proper credit be given"): U.S. Geological Survey, National Water Information System (NWIS), instantaneous streamflow values
+- Wikipedia page views (extended web traffic): None required (CC0); credit: Wikimedia
+- Wikipedia page views (rolling): None required (CC0); credit: Wikimedia
+- Wikipedia page views (weekly web traffic): None required (CC0); credit: Wikimedia

README.md

@@ -0,0 +1,140 @@
+---
+license: apache-2.0
+library_name: hirondel
+pipeline_tag: time-series-forecasting
+tags:
+  - time-series
+  - forecasting
+  - foundation-model
+  - zero-shot
+  - probabilistic-forecasting
+model-index:
+  - name: Hirondel-7M
+    results:
+      - task: {type: time-series-forecasting, name: Probabilistic forecasting}
+        dataset: {type: Salesforce/GiftEval, name: GIFT-Eval (97 tasks)}
+        metrics:
+          - {type: mase, name: MASE (normalised by Seasonal Naive), value: 0.7003}
+          - {type: crps, name: CRPS (normalised by Seasonal Naive), value: 0.4816}
+      - task: {type: time-series-forecasting, name: Probabilistic forecasting}
+        dataset: {type: autogluon/fev_datasets, name: fev-bench (100 tasks)}
+        metrics:
+          - {type: skill-score, name: Skill score, MASE (%), value: 32.91}
+          - {type: skill-score, name: Skill score, SQL (%), value: 44.90}
+      - task: {type: time-series-forecasting, name: Probabilistic forecasting}
+        dataset: {type: Real-TSF/TIME, name: TIME (98 tasks)}
+        metrics:
+          - {type: mase, name: MASE (normalised by Seasonal Naive), value: 0.6551}
+          - {type: crps, name: CRPS (normalised by Seasonal Naive), value: 0.5520}
+---
+
+![Hirondel](showcase/figures/banner.png)
+
+# Hirondel-7M: zero-shot probabilistic time-series forecasting
+
+*Hirondel — from French* hirondelle *(swallow), the bird said to tell the weather.*
+
+Hirondel is a 7M-parameter time-series forecasting model for zero-shot probabilistic forecasting: 9 quantiles from 0.1
+to 0.9, any horizon, CPU-friendly. Code and weights under Apache-2.0: https://github.com/tncrd/hirondel.
+
+**One of the most accurate small time-series foundation models.** At release (October 2026), it was the most accurate on GIFT-Eval, fev-bench and TIME among models under 10M parameters with public weights and a permissive license (official leaderboards, read on 2026-10-23; how this was checked: `showcase/README.md`).
+
+## Results
+
+<!-- results:start -->
+| Benchmark | MASE | CRPS (fev-bench: SQL) |
+|---|---|---|
+| GIFT-Eval (97 tasks, normalised by Seasonal Naive; lower is better) | 0.7003 | 0.4816 |
+| fev-bench (100 tasks, skill score in %; higher is better) | 32.91 | 44.90 |
+| TIME (98 tasks, normalised by Seasonal Naive; lower is better) | 0.6551 | 0.5520 |
+<!-- results:end -->
+
+These numbers are computed by us and not yet listed on the leaderboards. Added to the fev-bench leaderboard, they would rank 7 of 33 (SQL) and 6 of 33 (MASE) among the models the site shows by default (2026-10-23).
+
+![Accuracy against size, univariate setting](showcase/figures/univariate.png)
+
+Univariate setting: every model forecasts each series alone, without covariates (competitors run by us in forced
+univariate mode with their public packages; TiRex, TimesFM-2.5, FlowState-r1.1 and Moirai-2.0 univariate by design). Pareto fronts over models with
+a permissive license, the others shown hollow.
+
+| Model | Parameters | License | GIFT-Eval MASE | GIFT-Eval CRPS | TIME MASE | TIME CRPS | fev-bench SQL skill |
+|---|---|---|---|---|---|---|---|
+| Toto-2.0-4m | 4.1M | Apache-2.0 | 0.7591 | 0.5269 | 0.6940 | 0.5850 | 40.04 |
+| Timestral-5M | 5.2M | Modified MIT (revenue cap) | 0.7102 | 0.4836 | 0.6571 | 0.5490 | 44.36 |
+| **Hirondel (ours)** | 7.5M | Apache-2.0 | 0.7003 | 0.4816 | 0.6551 | 0.5520 | 44.90 |
+| Moirai-2.0-small | 11.4M | CC-BY-NC-4.0 | 0.7281 | 0.5164 | 0.7030 | 0.5885 | 39.31 |
+| FlowState-r1.1 | 18.5M | Apache-2.0 | 0.7014 | 0.4901 |  |  |  |
+| Toto-2.0-22m | 21.9M | Apache-2.0 | 0.7226 | 0.5009 | 0.6751 | 0.5654 | 42.37 |
+| Timestral-28M | 27.8M | Modified MIT (revenue cap) | 0.6817 | 0.4641 | 0.6455 | 0.5408 | 45.86 |
+| TiRex | 35.0M | NX-AI Community License | 0.7213 | 0.4921 | 0.6831 | 0.5731 | 42.58 |
+| Chronos-2 | 119.5M | Apache-2.0 | 0.7037 | 0.4911 | 0.6659 | 0.5584 | 43.97 |
+| TimesFM-2.5 | 231.3M | Apache-2.0 | 0.7050 | 0.4903 | 0.6686 | 0.5674 | 42.20 |
+
+### Industrial data: IND-Eval
+
+![IND-Eval, accuracy against size](showcase/figures/indeval.png)
+
+Five industrial sources of [IND-Eval](https://arxiv.org/abs/2610.16652), none in the training data. Hirondel: our run,
+each channel alone, no covariates. Other models: numbers published in the IND-Eval paper (joint forecasting, PRONTO's
+covariates). nMASE / nCRPS, lower is better:
+
+| Source | Hirondel (7.5M, ours) | Chronos-2 (119M) | TiRex-2 (83M) | TimesFM-3 (331M) |
+|---|---|---|---|---|
+| NEATS (offshore drilling) | 0.937 / 0.496 | 0.915 / 0.516 | **0.899** / 0.551 | 0.924 / **0.494** |
+| 3W (oil wells) | **0.754** / **0.653** | 0.847 / 0.749 | 0.837 / 0.786 | 0.756 / 0.690 |
+| PRONTO (multiphase flow) | 0.956 / 0.853 | 0.754 / 0.569 | 0.957 / 0.719 | **0.617** / **0.473** |
+| Mesa del Sol (solar and battery microgrid) | **0.838** / 0.493 | 0.864 / 0.948 | 0.901 / 0.542 | 0.841 / **0.481** |
+| Tablet compression (pharmaceutical) | 0.737 / 0.325 | 0.744 / 0.324 | 0.736 / 0.321 | **0.729** / **0.318** |
+| **Geometric mean, 5 sources** | 0.839 / 0.536 | 0.822 / 0.583 | 0.863 / 0.558 | **0.766** / **0.477** |
+
+Hirondel loses ground mainly on PRONTO, a process driven by control signals: the other models read these signals as
+covariates, Hirondel, being univariate, does not, so it cannot anticipate the changes they cause. On the four other
+sources it is within 5 % of the best model or ahead of it, with 11 to 47 times fewer parameters.
+
+![Accuracy against size, official leaderboards](showcase/figures/pareto.png)
+
+Official leaderboards, read on 2026-10-23 (some models use covariates or joint multivariate forecasting there).
+Sources: `showcase/data/` in the code repository.
+
+## Usage
+
+```python
+import numpy as np
+from hirondel import HirondelForecaster
+
+forecaster = HirondelForecaster.from_pretrained("Tncrd/Hirondel-7M")
+t = np.arange(1000)
+context = np.sin(2 * np.pi * t / 24) + 0.1 * np.random.default_rng(0).normal(size=t.size)
+forecast = forecaster.forecast(context, prediction_length=48)
+print(forecast.quantiles.shape)   # (1, 48, 9)
+```
+
+Install with `pip install hirondel`. Inference details, tail levels, hardware and numerical parity: see the code
+repository's README (https://github.com/tncrd/hirondel).
+
+## Training data
+
+Real data under licenses that allow this release, plus synthetic data. Every source, its license, its attribution and
+the changes made are listed in `DATA.md`; `NOTICE` gathers the required attributions. No dataset of GIFT-Eval's test
+split, fev-bench or TIME is in the training data.
+
+## Limitations
+
+- Univariate: no covariates, no cross-series information.
+- Context: 2 048 points at full resolution plus a coarse reading of up to 8 192 points.
+- Tail levels 0.01 / 0.05 / 0.95 / 0.99 are extrapolated, not learned.
+
+## License
+
+Apache-2.0 for code and weights. `NOTICE` must be kept when redistributing.
+
+## Citation
+
+```bibtex
+@misc{hirondel2026,
+  title        = {Hirondel-7M: zero-shot probabilistic time-series forecasting},
+  author       = {{The Hirondel Author (Tncrd)}},
+  year         = {2026},
+  howpublished = {\url{https://huggingface.co/Tncrd/Hirondel-7M}}
+}
+```

config.json

@@ -0,0 +1,54 @@
+{
+  "format_version": 1,
+  "d_model": 360,
+  "n_heads": 4,
+  "n_layers": 6,
+  "layer_types": [
+    "delta",
+    "attention",
+    "delta",
+    "attention",
+    "delta",
+    "attention"
+  ],
+  "ffn_hidden": 480,
+  "embed_hidden": 240,
+  "patch_size": 16,
+  "conv_kernel": 4,
+  "delta_write_max": 2.0,
+  "delta_chunk": 16,
+  "norm_eps": 1e-06,
+  "rope_base": 10000.0,
+  "time_scale": 8.0,
+  "loop_passes": 2,
+  "anchor_rank": 64,
+  "quantile_levels": [
+    0.1,
+    0.2,
+    0.3,
+    0.4,
+    0.5,
+    0.6,
+    0.7,
+    0.8,
+    0.9
+  ],
+  "context_length": 2048,
+  "native_horizon": 720,
+  "coarse_factors": [
+    4
+  ],
+  "fusion": {
+    "state_dim": 32,
+    "spectral_bands": 16,
+    "basis_size": 8,
+    "hidden": 128,
+    "max_horizon": 720
+  },
+  "input_transform": "asinh",
+  "output_clip": 15.0,
+  "min_scale": 1e-05,
+  "range_margin": 10000.0,
+  "flip": true,
+  "clip_nonnegative": true
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddfc724f5d64df6651c718a88c660fa1dd6512b49deb44fc6a2009ed01a91a32
+size 29863660