batched Inference with Different Frequencies

[e-hossam96]

Hello All

Description

We need a way to handle batch inference with different frequencies. I searched the entire documentation and I couldn't point my finger on it.

To Reproduce

This is the current example function I use. I pass the Dataset a list of data frames, each with the "target" column and "datetime" index.

def get_forecast(
    data: pd.DataFrame | list[pd.DataFrame],
    forecaster: PyTorchPredictor | ProphetPredictor,
) -> list[np.ndarray]:
    torch.manual_seed(forecasting_settings.seed)
    seed_everything(forecasting_settings.seed)
    results = list(forecaster.predict(dataset=PandasDataset(data)))
    demands = [np.median(result.samples, axis=0) for result in results]
    return demands

Error message or code output

The code works fine but it takes so long to complete.

Environment

• Win 11
• Python version: 3.12.9
• GluonTS version: gluonts[prophet,torch]==0.14.4
• MXNet version: ...