Method sparse

MinkowskiEngine/MinkowskiSparseTensor.py:344–455

r"""Convert the :attr:`MinkowskiEngine.SparseTensor` to a torch sparse tensor. Args: :attr:`min_coords` (torch.IntTensor, optional): The min coordinates of the output sparse tensor. Must be divisible by the current :attr:`tensor_stride`.

(self, min_coords=None, max_coords=None, contract_coords=True)

    # Conversion functions
    def sparse(self, min_coords=None, max_coords=None, contract_coords=True):
        r"""Convert the :attr:`MinkowskiEngine.SparseTensor` to a torch sparse
        tensor.

        Args:
            :attr:`min_coords` (torch.IntTensor, optional): The min
            coordinates of the output sparse tensor. Must be divisible by the
            current :attr:`tensor_stride`.

            :attr:`max_coords` (torch.IntTensor, optional): The max coordinates
            of the output sparse tensor (inclusive). Must be divisible by the
            current :attr:`tensor_stride`.

            :attr:`contract_coords` (bool, optional): Given True, the output
            coordinates will be divided by the tensor stride to make features
            contiguous.

        Returns:
            :attr:`spare_tensor` (torch.sparse.Tensor): the torch sparse tensor
            representation of the self in `[Batch Dim, Spatial Dims..., Feature
            Dim]`. The coordinate of each feature can be accessed via
            `min_coord + tensor_stride * [the coordinate of the dense tensor]`.

            :attr:`min_coords` (torch.IntTensor): the D-dimensional vector
            defining the minimum coordinate of the output sparse tensor. If
            :attr:`contract_coords` is True, the :attr:`min_coords` will also
            be contracted.

            :attr:`tensor_stride` (torch.IntTensor): the D-dimensional vector
            defining the stride between tensor elements.

        """

        if min_coords is not None:
            assert isinstance(min_coords, torch.IntTensor)
            assert min_coords.numel() == self._D
        if max_coords is not None:
            assert isinstance(max_coords, torch.IntTensor)
            assert min_coords.numel() == self._D

        def torch_sparse_Tensor(coords, feats, size=None):
            if size is None:
                if feats.dtype == torch.float64:
                    return torch.sparse.DoubleTensor(coords, feats)
                elif feats.dtype == torch.float32:
                    return torch.sparse.FloatTensor(coords, feats)
                else:
                    raise ValueError("Feature type not supported.")
            else:
                if feats.dtype == torch.float64:
                    return torch.sparse.DoubleTensor(coords, feats, size)
                elif feats.dtype == torch.float32:
                    return torch.sparse.FloatTensor(coords, feats, size)
                else:
                    raise ValueError("Feature type not supported.")

        # Use int tensor for all operations
        tensor_stride = torch.IntTensor(self.tensor_stride)