nandivsum

nandivsum(data: Tensor, dim: tuple[int, ...] | int | None = None, keepdim: bool = False, denom: int | float | complex = 1, **kwargs: Any) → Tensor[source]

Returns a tensor with dimensions reduced via summation then divided by a constant, excluding NaN values.

This is a wrapper around torch.nansum().

Parameters:

  • data (torch.Tensor) – tensor to which operations should be applied.

  • dim (tuple[int, ...] | int | None, optional) – dimension(s) along which the reduction should be applied, all dimensions when None. Defaults to None.

  • keepdim (bool, optional) – if the dimensions should be retained in the output. Defaults to False.

  • denom (int | float | complex, optional) – value by which to divide the sum. Defaults to 1.

Returns:

dimensionally reduced tensor.

Return type:

torch.Tensor

Tip

This is useful in cases where the mean over a larger sample is to be computed, but only a subset of the sample is being reduced.