spine.data.batch.TensorBatch

class spine.data.batch.TensorBatch(data: Any | SparseTensorLike, counts: Sequence[int] | Any | None = None, batch_size: int | None = None, is_sparse: bool = False, has_batch_col: bool = False, coord_cols: Sequence[int] | ndarray | None = None)[source]

Batched tensor with the necessary methods to slice it.

Attributes:
batch_ids

Returns the batch ID of each of the elements in the tensor.

shape

Shape of the underlying data.

splits

Boundaries needed to split the data into its constituents.

tensor

Alias for the underlying data stored.

Methods

apply_mask(mask)

Apply a global mask to the underlying tensor, update batching.

from_list(data_list)

Builds a batch from a list of tensors.

get_counts(batch_ids, batch_size)

Finds the number of elements in each entry, provided a batch ID list.

get_edges(counts)

Finds the edges between successive entries in the batch.

merge(tensor_batch)

Merge this tensor batch with another.

split()

Breaks up the tensor batch into its constituents.

to_cm(meta)

Converts the pixel coordinates of the tensor to cm.

to_numpy()

Cast underlying tensor to a np.ndarray and return a new instance.

to_px(meta)

Converts the coordinates of the tensor to pixel indexes.

to_tensor([dtype, device])

Cast underlying tensor to a torch.tensor and return a new instance.
__init__(data: Any | SparseTensorLike, counts: Sequence[int] | Any | None = None, batch_size: int | None = None, is_sparse: bool = False, has_batch_col: bool = False, coord_cols: Sequence[int] | ndarray | None = None) → None[source]

Initialize the attributes of the class.

Parameters:

  • data (Union[np.ndarray, torch.Tensor, ME.SparseTensor]) – (N, C) Batched tensors

  • counts (Union[List[int], np.ndarray, torch.Tensor]) –

    1. Number of data rows in each entry

  • batch_size (int, optional) – Number of entries that make up the batched data

  • is_sparse (bool, default False) – If initializing from an ME sparse data, flip to True

  • has_batch_col (bool, default False) – Wheather the tensor has a column specifying the batch ID

  • coord_cols (Union[List[int], np.ndarray], optional) – List of columns specifying coordinates

Methods

__init__(data[, counts, batch_size, ...])

Initialize the attributes of the class.

apply_mask(mask)

Apply a global mask to the underlying tensor, update batching.

from_list(data_list)

Builds a batch from a list of tensors.

get_counts(batch_ids, batch_size)

Finds the number of elements in each entry, provided a batch ID list.

get_edges(counts)

Finds the edges between successive entries in the batch.

merge(tensor_batch)

Merge this tensor batch with another.

split()

Breaks up the tensor batch into its constituents.

to_cm(meta)

Converts the pixel coordinates of the tensor to cm.

to_numpy()

Cast underlying tensor to a np.ndarray and return a new instance.

to_px(meta)

Converts the coordinates of the tensor to pixel indexes.

to_tensor([dtype, device])

Cast underlying tensor to a torch.tensor and return a new instance.

Attributes

batch_ids

Returns the batch ID of each of the elements in the tensor.

shape

Shape of the underlying data.

splits

Boundaries needed to split the data into its constituents.

tensor

Alias for the underlying data stored.

data

counts

edges

batch_size

has_batch_col

coord_cols

data: Any | SparseTensorLike
counts: Any
edges: Any
batch_size: int
has_batch_col: bool
coord_cols: Sequence[int] | ndarray | None
property tensor: Any | SparseTensorLike

Alias for the underlying data stored.

Returns:

Underlying tensor of data

Return type:

Union[np.ndarray, torch.Tensor, ME.SparseTensor]

property batch_ids: Any

Returns the batch ID of each of the elements in the tensor.

Returns:

  1. Batch ID of each element in the tensor

Return type:

Union[np.ndarray, torch.Tensor]

split() → list[Any | SparseTensorLike][source]

Breaks up the tensor batch into its constituents.

Returns:

List of one tensor per entry in the batch

Return type:

List[Union[np.ndarray, torch.Tensor]]

apply_mask(mask: Any) → None[source]

Apply a global mask to the underlying tensor, update batching.

Parameters:

mask (Union[np.ndarray, torch.Tensor]) –

  1. Boolean mask to apply to the underlying tensor

merge(tensor_batch: TensorBatch) → TensorBatch[source]

Merge this tensor batch with another.

Parameters:

tensor_batch (TensorBatch) – Other tensor batch object to merge with

Returns:

Merged tensor batch

Return type:

TensorBatch

to_numpy() → TensorBatch[source]

Cast underlying tensor to a np.ndarray and return a new instance.

Returns:

New TensorBatch object with an underlying np.ndarray tensor.

Return type:

TensorBatch

to_tensor(dtype: Any = None, device: Any = None) → TensorBatch[source]

Cast underlying tensor to a torch.tensor and return a new instance.

Parameters:

  • dtype (torch.dtype, optional) – Data type of the tensor to create

  • device (torch.device, optional) – Device on which to put the tensor

Returns:

New TensorBatch object with an underlying np.ndarray tensor.

Return type:

TensorBatch

to_cm(meta: Any) → None[source]

Converts the pixel coordinates of the tensor to cm.

Parameters:

meta (Meta) – Metadata information about the rasterized image

to_px(meta: Any) → None[source]

Converts the coordinates of the tensor to pixel indexes.

Parameters:

meta (Meta) – Metadata information about the rasterized image

classmethod from_list(data_list: Sequence[Any]) → TensorBatch[source]

Builds a batch from a list of tensors.

Parameters:

data_list (List[Union[np.ndarray, torch.Tensor]]) – List of tensors, exactly one per batch