MMM.forward_pass#

MMM.forward_pass(x, dims)[source]#

Transform channel input into target contributions of each channel.

This method handles the ordering of the adstock and saturation transformations.

This method must be called from without a pm.Model context but not necessarily in the instance’s model. A dim named “channel” is required associated with the number of columns of x.

Parameters:

xpt.TensorVariable | npt.NDArray[np.float64]: The channel input which could be spends or impressions

Returns:

The contributions associated with the channel input

Examples

>>> mmm = MMM(
    date_column="date",
    channel_columns=["channel_1", "channel_2"],
    target_column="target",
)