Falcon SDK

The official Python SDK for the Falcon-TST forecasting API. Falcon-TST is a family of large-scale time-series foundation models developed by Ant International.

Install

pip install falcon-tst

Usage

import numpy as np

from falcontst import FalconClient

client = FalconClient()

result = client.quantile_predict(
    context=np.array([
        [1.2, 3.4, 5.6, 7.8, 9.0],
        [2.1, 4.3, 6.5, 8.7, 0.9],
    ]),
    prediction_length=3,
    model_name="demo_model",
    group_ids=np.array([0, 0]),
    input_mask=np.array([
        [1, 1, 1, 1, 1],
        [1, 1, 1, 0, 0],
    ]),
    is_multivariate=False,
)

print(result)

Batch prediction uses the same object fields as quantile_predict, but sends multiple objects to the batch endpoint:

result = client.batch_predict(
    [
        {
            "context": np.array([[1.2, 3.4, 5.6, 7.8, 9.0]]),
            "prediction_length": 3,
            "model_name": "demo_model",
            "group_ids": np.array([0]),
            "input_mask": np.array([[1, 1, 1, 1, 1]]),
            "is_multivariate": False,
        },
        {
            "context": np.array([[2.1, 4.3, 6.5, 8.7, 0.9]]),
            "prediction_length": 3,
        },
    ]
)

Each context should be a two-dimensional numpy.ndarray. For a single time series, use np.array([[1.2, 3.4, 5.6]]) instead of np.array([1.2, 3.4, 5.6]).