h2oai/timeh2o-v1-tiny

TimeH2O Tiny

Time series foundation model by H2O.ai.

Usage

pip install h2ofm

import numpy as np
from h2ofm import TimeH2O

model = TimeH2O.from_pretrained("h2oai/timeh2o-v1-tiny")
forecast = model.forecast(np.sin(np.arange(200) / 10), horizon=24)

forecast.point  # (24,), median forecast; also forecast.median
forecast.quantiles  # (24, 9), default levels 0.1 through 0.9
forecast.quantile(0.9)  # (24,)

License

Non-commercial. See LICENSE.

Training Data

We pretrain on data from the following public sources. Datasets overlapping with fev-bench, GIFT-Eval, M4, and QuitoBench were excluded from training at the dataset or test-split level.

• autogluon/chronos_datasets (per-dataset licenses, see dataset card)
• Salesforce/GiftEvalPretrain (Apache-2.0)
• Salesforce/lotsa_data (Apache-2.0)
• hq-bench/quito-corpus (CC-BY 4.0)
• Dingdong-Inc/FreshRetailNet-LT (CC-BY 4.0)
• Azure Functions 2019 trace (CC-BY 4.0)
• Iowa Liquor Sales (Iowa Open Data terms)
• BLS Average Prices (public domain)