JuaAI/ts-icl-pretraining-corpus

TS-ICL Pretraining Corpus (community reconstruction)

A unified, cleaned reconstruction of the univariate pretraining corpus described in Table 5 of TS-ICL: A Flexible Time-Indexed Foundation Model for Time Series via In-Context Learning (Le Naour, Nabil & Petralia, EDF R&D; arXiv:2606.05878). The TS-ICL authors did not release their pretraining data pipeline, so this corpus is rebuilt from the named upstream sources (LOTSA, Chronos, and the TempoPFN synthetic generators) and normalized into a single schema.

Curated by: Jua — reconstruction, normalization, synthetic generation, validation and packaging.

Not an official release. This is a derivative dataset built by Jua for research reproducibility; it is not affiliated with or endorsed by the TS-ICL authors. Each subset retains the license of its upstream source (see Licensing).

At a glance

• 39 datasets, 1,399,719 series, 8,690,897,054 observations
• Sources: LOTSA (21), Chronos (10), TempoPFN synthetic (8)
• Single consistent schema, sharded Parquet (zstd)

Repository layout

data/<config>/*.parquet     # one folder per dataset (= HF config)
meta/datasets.yaml          # build registry (sources, fingerprints, subsampling)
meta/reports/<config>.json  # per-dataset build/QA report
meta/qa.json                # aggregated QA
QUALITY_REPORT.md           # human-readable QA table

Schema

Every row is one (possibly multivariate) time series:

field	type	description
item_id	string	unique id within a dataset
dataset	string	dataset name (= config name)
domain	string	Energy / Climate / Traffic / Cloud / Web / Econ-Fin / Health / Synthetic
source	string	lotsa / chronos / chronos_extra / tempopfn
freq	string	pandas offset alias (null for synthetic)
start	timestamp[s]	start time (null for synthetic)
target	list<list>	values as [num_channels, length] (univariate = [1, length])
num_channels	int32	number of channels
length	int32	series length
weight	float32	Table-5 sampling coefficient (training-sampler metadata)

How to load

from datasets import load_dataset

# one dataset (config)
ds = load_dataset("<repo_id>", "nn5_weekly", split="train")

# everything, streamed
ds = load_dataset("<repo_id>", "all", split="train", streaming=True)

Datasets (provenance)

config	paper name	source	upstream location	domain	freq	weight	series
bdg2_bull	BDG-2 Bull	lotsa	bull	Energy	H	25	41
bdg2_fox	BDG-2 Fox	lotsa	bdg-2_fox	Energy	H	5	135
bdg2_panther	BDG-2 Panther	lotsa	bdg-2_panther	Energy	H	2.5	105
buildings_900k	BuildingsBench900k	lotsa	buildings_900k	Energy	H	0.02048	100,000
residential_load_power	Residential Load Power	lotsa	residential_load_power	Energy	1T	1.2	271
residential_pv_power	Residential PV Power	lotsa	residential_pv_power	Energy	1T	1.5	233
china_air_quality	China Air Quality	lotsa	china_air_quality	Climate	H	0.3	437
cmip6_2000	CMIP6 2000	lotsa	cmip6_2000	Climate	6H	0.057	8,192
era5_1989	ERA5 1989	lotsa	era5_1989	Climate	H	0.085	8,192
era5_1990	ERA5 1990	lotsa	era5_1990	Climate	H	0.085	8,192
era5_1991	ERA5 1991	lotsa	era5_1991	Climate	H	0.085	8,192
subseasonal	Subseasonal	lotsa	subseasonal	Climate	1D	0.3	862
subseasonal_precip	Subseasonal Precipitation	lotsa	subseasonal_precip	Climate	1D	1.2	862
pems04	PEMS04	lotsa	PEMS04	Traffic	5T	1.2	307
pems07	PEMS07	lotsa	PEMS07	Traffic	5T	1.2	883
pems08	PEMS08	lotsa	PEMS08	Traffic	5T	2.1	170
q_traffic	Q-TRAFFIC	lotsa	Q-TRAFFIC	Traffic	15T	0.024	45,148
alibaba_cluster_trace_2018	Alibaba Cluster Trace 2018	lotsa	alibaba_cluster_trace_2018	Cloud	5T	0.009	58,409
monash_m3_monthly	Monash M3 Monthly	lotsa	monash_m3_monthly	Econ/Fin	M	0.72	1,428
nn5_weekly	NN5 Weekly	lotsa	nn5_weekly	Econ/Fin	W	5	111
project_tycho	Project Tycho	lotsa	project_tycho	Health	W	0.21	1,258
australian_electricity	Australian Electricity	chronos	monash_australian_electricity	Energy	30T	220	5
wind_farms_hourly	Wind Farms H	chronos	wind_farms_hourly	Energy	H	4	337
wind_farms_daily	Wind Farms D	chronos	wind_farms_daily	Energy	D	2	337
weatherbench_daily	Weatherbench daily	chronos	weatherbench_daily	Climate	1D	0.1024	10,000
mexico_city_bikes	Mexico City Bikes	chronos	mexico_city_bikes	Traffic	H	2.5	494
taxi_30min	Taxi (30 Min.)	chronos	taxi_30min	Traffic	30T	0.88	2,428
taxi_1h	Taxi (Hourly)	chronos	taxi_1h	Traffic	H	0.88	2,428
uber_tlc_hourly	Uber TLC (Hourly)	chronos	uber_tlc_hourly	Traffic	H	4	262
wiki_daily_100k	Wiki Daily	chronos	wiki_daily_100k	Web	D	0.00512	100,000
kernel_synth_1m	Kernel Synth 1M	chronos	training_corpus/kernel_synth_1m	Synthetic	None	0.001024	1,000,000
syn_anomaly	Anomaly	tempopfn	AnomalyGenerator	Synthetic	None	0.0256	5,000
syn_forecastpfn	ForecastPFN	tempopfn	ForecastPFNGenerator	Synthetic	None	1	5,000
syn_gp	GP	tempopfn	GPGenerator	Synthetic	None	0.4096	5,000
syn_sawtooth	Sawtooth	tempopfn	SawToothGenerator	Synthetic	None	0.0512	5,000
syn_sinewave	Sinewave	tempopfn	SineWaveGenerator	Synthetic	None	0.1024	5,000
syn_spikes	Spikes	tempopfn	SpikesGenerator	Synthetic	None	0.0256	5,000
syn_step	Step	tempopfn	StepGenerator	Synthetic	None	0.0512	5,000
syn_ou	OU	tempopfn	OrnsteinUhlenbeckProcessGenerator	Synthetic	None	0.4096	5,000

Reconstruction methodology

• LOTSA subsets: from Salesforce/lotsa_data (Arrow); targets passed through.
• Chronos subsets: from autogluon/chronos_datasets; per-series timestamp + value columns converted to start / freq (inferred) / target. KernelSynth-1M from training_corpus/kernel_synth_1m.
• Spanish Weather: from autogluon/chronos_datasets_extra (Kaggle-backed; needs Kaggle credentials), reshaped to 5 cities x {temp, pressure, humidity}. (Omitted unless built with Kaggle creds.)
• Synthetic (syn_*): regenerated with the open-source automl/TempoPFN generators (Anomaly, ForecastPFN, GP, Sawtooth, SineWave, Spikes, Step, OU), 5,000 series each, fixed per-dataset seeds. Exact paper samples are not recoverable; these are statistically equivalent and reproducible from the documented seeds.
• Paper-faithful downsampling (seeded): buildings_900k -> 100k of ~1.8M series; weatherbench_daily -> 10k; era5_* -> 15 of 45 channels; cmip6_2000 -> 22 of 53.
• Cleaning: float32; inf -> NaN (missing values preserved as NaN); empty/all-NaN series dropped; frequency aliases normalized.
• Validation: #series and #channels checked against Table 5 (hard); max_length is informational (current upstream snapshots sometimes have longer raw series than Table 5).

Known deviations from the paper

• syn_gp uses length 2048 instead of 10,000. Exact GP at length 10,000 is computationally infeasible (Cholesky fails -> symeig on 10k x 10k matrices, ~minutes each; ~24 h for 5,000 series). 2048 is a standard GP/KernelSynth prior length.
• A few datasets (e.g. BDG-2, wind_farms_hourly) have longer max_length than Table 5 because the current upstream snapshot contains longer raw series. Values are unmodified.

Licensing

Derivative reconstruction; each subset is governed by its upstream license:

• LOTSA subsets: see Salesforce/lotsa_data (per-dataset licenses).
• Chronos subsets: see autogluon/chronos_datasets.
• Spanish Weather: Kaggle "energy-consumption-generation-prices-and-weather".
• Synthetic syn_*: generated with automl/TempoPFN (Apache-2.0).

Users must comply with the original licenses and cite the original dataset authors.

Citation

If you use this corpus, please cite both this dataset and the upstream sources.

This dataset (the reconstruction):

@misc{jua2026tsiclcorpus,
  title        = {TS-ICL Pretraining Corpus (reconstruction)},
  author       = {Jua},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/JuaAI/ts-icl-pretraining-corpus}},
  note         = {Community reconstruction of the TS-ICL (arXiv:2606.05878) pretraining mix}
}

Upstream sources:

@article{lenaour2026tsicl,
  title  = {TS-ICL: A Flexible Time-Indexed Foundation Model for Time Series via In-Context Learning},
  author = {Le Naour, Etienne and Nabil, Tahar and Petralia, Adrien},
  journal= {arXiv preprint arXiv:2606.05878},
  year   = {2026}
}
@article{woo2024moirai, title={Unified Training of Universal Time Series Forecasting Transformers (LOTSA)}, author={Woo, Gerald and others}, year={2024}}
@article{ansari2024chronos, title={Chronos: Learning the Language of Time Series}, author={Ansari, Abdul Fatir and others}, journal={arXiv:2403.07815}, year={2024}}
@misc{moroshan2025tempopfn, title={TempoPFN: Synthetic Pre-training of Linear RNNs for Zero-Shot Time Series Forecasting}, author={Moroshan, Vladyslav and Siems, Julien and Zela, Arber and Carstensen, Timur and Hutter, Frank}, eprint={2510.25502}, year={2025}}