Pipeline

Typical workflow:

1. Prepare data with stella.FlareDataSet.
2. Initialize stella.ConvNN with desired backend.
3. Train with cnn.train_models(...) or predict with packaged models.

Example prediction using packaged model:

from stella.neural_network import ConvNN
from stella.models import get_model_path

cnn = ConvNN(output_dir="./results")
cnn.predict(
    modelname=get_model_path(),
    times=lc.time.value,
    fluxes=lc.flux.value,
    errs=getattr(lc, "flux_err", None).value if getattr(lc, "flux_err", None) is not None else np.zeros_like(lc.time.value),
)