sktime/chronos2-onnx

Model

0

stars

5

commits

2

linked in READMEs

Sep 1, 2026

updated

forecasting
fp32
onnx
onnxruntime
time-series
zero-shot
Browse cluster: Time Series Forecasting with Transformers

README

Chronos-2 (fp32 ONNX)

Unquantized fp32 ONNX export of amazon/chronos-2 (Apache-2.0).

fp32 output verified within 0.0001% of forecast spread of the official Chronos2Pipeline. Unofficial export, not affiliated with or endorsed by the model authors.

Contract

inputdtypeshapenotes
contextfloat32['batch', 2048]NaN = missing; left-pad short series with NaN
group_idsint64['batch']rows sharing an id are forecast jointly (multivariate); distinct ids = independent

Output quantiles: float32 ['batch', 64, 21]. Quantile levels: [0.01, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 0.99].

One forward pass. Left-pad short series with NaN (the model is missing-aware; padding is exactly equivalent to the natural-length call). Median is level index 10.

See manifest.json for the machine-readable contract and the artifact's sha256.

Contributors

TobiasPitters

5 commits

sktime/chronos2-onnx

Model

0

stars

5

commits

2

linked in READMEs

Sep 1, 2026

updated

forecasting
fp32
onnx
onnxruntime
time-series
zero-shot
Browse cluster: Time Series Forecasting with Transformers

README

Chronos-2 (fp32 ONNX)

Unquantized fp32 ONNX export of amazon/chronos-2 (Apache-2.0).

fp32 output verified within 0.0001% of forecast spread of the official Chronos2Pipeline. Unofficial export, not affiliated with or endorsed by the model authors.

Contract

inputdtypeshapenotes
contextfloat32['batch', 2048]NaN = missing; left-pad short series with NaN
group_idsint64['batch']rows sharing an id are forecast jointly (multivariate); distinct ids = independent

Output quantiles: float32 ['batch', 64, 21]. Quantile levels: [0.01, 0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95, 0.99].

One forward pass. Left-pad short series with NaN (the model is missing-aware; padding is exactly equivalent to the natural-length call). Median is level index 10.

See manifest.json for the machine-readable contract and the artifact's sha256.

Contributors

TobiasPitters

5 commits