config

Run-level configuration for torchcrop simulations.

RunConfig dataclass

Simulation run configuration.

Attributes:

Name	Type	Description
n_days	int	Number of daily time steps T in the simulation.
start_doy	int	Day-of-year of the first simulated day. Used by the astronomical component to compute solar declination and daylength.
dt	float	Integration step in days. Defaults to 1.0 — the Lintul5 model is formulated for daily forward-Euler integration.
smooth	bool	If True process modules use smooth (sigmoid-blend) replacements for stage-based branching; otherwise torch.where is used.
dtype	str	Tensor dtype for the simulation ("float32" or "float64"). Gradient checks require float64.
device	str	Device string ("cpu", "cuda", ...).

Source code in torchcrop/config.py

@dataclass
class RunConfig:
    """Simulation run configuration.

    Attributes:
        n_days: Number of daily time steps ``T`` in the simulation.
        start_doy: Day-of-year of the first simulated day. Used by the
            astronomical component to compute solar declination and daylength.
        dt: Integration step in days. Defaults to ``1.0`` — the Lintul5 model
            is formulated for daily forward-Euler integration.
        smooth: If ``True`` process modules use smooth (sigmoid-blend)
            replacements for stage-based branching; otherwise ``torch.where``
            is used.
        dtype: Tensor dtype for the simulation (``"float32"`` or
            ``"float64"``). Gradient checks require ``float64``.
        device: Device string (``"cpu"``, ``"cuda"``, ...).
    """

    n_days: int = 200
    start_doy: int = 1
    dt: float = 1.0
    smooth: bool = False
    dtype: str = "float32"
    device: str = "cpu"