t0-gamma ONNX checkpoints
ONNX export of theforecastingcompany/t0-gamma for
OpenSTEF, produced by
openstef-checkpoints.
This export targets a zero-shot time-series forecasting model from the
Forecasting Company's t0 family. It is written for a fixed context and horizon window
with a dynamic batch axis, so it can be verified against the upstream tfc-t0
reference implementation and published as a portable ONNX checkpoint.
Inputs
| Tensor | Shape | Meaning |
|---|---|---|
past_values |
(batch, context_length) |
Load measurements before the origin |
future_covariates |
(batch, horizon_length, n_covariates) |
Weather known into the forecast window |
quantile_predictions |
(batch, horizon_length, n_quantiles) |
The model's native quantile grid |
Covariates must be point-in-time: the values that existed at the origin, not the
values observed afterwards. The graph carries a fixed budget of covariate slots. Fill a
slot you have no data for with NaN in both past_covariates and future_covariates
and the graph blanks it before the model runs, so any covariate count up to the budget
fits this one export.
Export notes
t0 guards every input with an if on a tensor reduction - row types, context width,
segment count, LocScale shapes, attention masking, quantile levels - and
torch.export cannot trace any of them. Under this contract each guard is statically
decidable, so each is replaced by the branch the contract implies. No model
mathematics is altered: the transformer, the scaler's statistics and the rollout are
the upstream code paths.
Why a re-export and not a first-party ONNX
The Forecasting Company publishes official ONNX exports for the earlier t0
releases (t0-beta-onnx-fp16,
t0-beta-onnx-int8,
t0-alpha-onnx-fp16),
but has not published one for t0-gamma at the time of this export. Even where
they exist, those artifacts do not fit OpenSTEF's checkpoint contract:
- Different graph interface. The official exports declare
"interface": "grouped": grouped multivariate inputs with no(batch, ...)axis, and a compute horizon that must be a positive multiple of 32, up to 1024. OpenSTEF's forecasters feedpast_values,future_covariatesandpast_covariatesas(batch, ...)arrays and read a(batch, horizon, n_quantiles)output, so the official graph is not a drop-in. - No fp32 weights. The official artifacts store large weights in reduced precision (fp16 / int8) and are tuned for browser, WASM and WebGPU execution. The checkpoints this project publishes are uniform fp32 with a dynamic batch.
- No OpenSTEF metadata. OpenSTEF's runtime reads context length, horizon, covariate count, output name and the native quantile grid from the checkpoint metadata. The official repos describe their own schema instead.
A first-party artifact could in principle be wrapped behind an adapter, but that would pin OpenSTEF to a quantized grouped-interface graph and sit outside this repository's trace, verify and publish pipeline. A clean fp32 re-export keeps the interface and precision uniform.
Variants
| File | Precision | Static shapes | Max deviation vs torch |
|---|---|---|---|
t0-gamma_static.onnx |
fp32 | yes | 4.005e-05 |
t0-gamma.onnx |
fp32 | no | 6.866e-05 |
Export window
- Context: 512 steps
- Horizon: 96 steps
- Resolution: 15 minute step
Provenance
- Source model:
theforecastingcompany/t0-gamma@unknown - Exporter:
openstef-checkpoints@1e2712e - Tooling: onnx=1.19.0 onnxruntime=1.27.0 torch=2.14.1 tfc-t0=0.6.0
- Exported: 2026-10-26T20:53:53+00:00
License and attribution
These ONNX checkpoints are derived from
theforecastingcompany/t0-gamma and released under the same
license, apache-2.0. Attribution and all rights to the model weights remain with the
upstream authors; if you use the model in research, please cite their work.