param_net

Parameter net — predict crop parameters from site/weather embeddings.

ParameterNet (Module)

MLP that maps a context embedding to a vector of crop parameters.

This is useful when you want to condition parameters on site/cultivar metadata (e.g., a learned embedding for each cultivar identifier).

Parameters:

Name	Type	Description	Default
embedding_dim	int	Size of the input context embedding.	required
n_params	int	Number of crop parameters to predict.	required
hidden_dim	int	Hidden-layer width.	64
n_hidden	int	Number of hidden layers (each followed by Tanh).	2

Source code in torchcrop/nn/param_net.py

class ParameterNet(nn.Module):
    """MLP that maps a context embedding to a vector of crop parameters.

    This is useful when you want to condition parameters on site/cultivar
    metadata (e.g., a learned embedding for each cultivar identifier).

    Args:
        embedding_dim: Size of the input context embedding.
        n_params: Number of crop parameters to predict.
        hidden_dim: Hidden-layer width.
        n_hidden: Number of hidden layers (each followed by ``Tanh``).
    """

    def __init__(
        self,
        embedding_dim: int,
        n_params: int,
        hidden_dim: int = 64,
        n_hidden: int = 2,
    ) -> None:
        super().__init__()
        layers: list[nn.Module] = []
        d = embedding_dim
        for _ in range(n_hidden):
            layers += [nn.Linear(d, hidden_dim), nn.Tanh()]
            d = hidden_dim
        layers.append(nn.Linear(d, n_params))
        self.mlp = nn.Sequential(*layers)

    def forward(self, embedding: torch.Tensor) -> torch.Tensor:
        """Predict crop parameters from a context embedding.

        Args:
            embedding: Context embedding tensor of shape
                ``[..., embedding_dim]``.

        Returns:
            Predicted parameter vector of shape ``[..., n_params]``.
        """
        return self.mlp(embedding)

forward(self, embedding)

Predict crop parameters from a context embedding.

Parameters:

Name	Type	Description	Default
embedding	torch.Tensor	Context embedding tensor of shape [..., embedding_dim].	required

Returns:

Type	Description
torch.Tensor	Predicted parameter vector of shape [..., n_params].

Source code in torchcrop/nn/param_net.py

def forward(self, embedding: torch.Tensor) -> torch.Tensor:
    """Predict crop parameters from a context embedding.

    Args:
        embedding: Context embedding tensor of shape
            ``[..., embedding_dim]``.

    Returns:
        Predicted parameter vector of shape ``[..., n_params]``.
    """
    return self.mlp(embedding)