Skip to contents

Data Descriptor

Source: R/DataDescriptor.R
DataDescriptor.Rd

A data descriptor is a rather internal data structure used in the lazy_tensor data type. In essence it is an annotated torch::dataset and a preprocessing graph (consisting mosty of PipeOpModule operators). The additional meta data (e.g. pointer, shapes) allows to preprocess lazy_tensors in an mlr3pipelines::Graph just like any (non-lazy) data types. The preprocessing is applied when materialize() is called on the lazy_tensor.

To create a data descriptor, you can also use the as_data_descriptor() function.

Details

While it would be more natural to define this as an S3 class, we opted for an R6 class to avoid the usual trouble of serializing S3 objects. If each row contained a DataDescriptor as an S3 class, this would copy the object when serializing.

See also

ModelDescriptor, lazy_tensor

Public fields

dataset

(torch::dataset)
The dataset.

graph

(Graph)
The preprocessing graph.

dataset_shapes

(named list() of (integer() or NULL))
The shapes of the output.

input_map

(character())
The input map from the dataset to the preprocessing graph.

pointer

(character(2))
The output pointer.

pointer_shape

(integer() | NULL)
The shape of the output indicated by pointer.

dataset_hash

(character(1))
Hash for the wrapped dataset.

hash

(character(1))
Hash for the data descriptor.

graph_input

(character())
The input channels of the preprocessing graph (cached to save time).

pointer_shape_predict

(integer() or NULL)
Internal use only.

Methods

Public methods

Method new()

Creates a new instance of this R6 class.

Usage

DataDescriptor$new(
  dataset,
  dataset_shapes = NULL,
  graph = NULL,
  input_map = NULL,
  pointer = NULL,
  pointer_shape = NULL,
  pointer_shape_predict = NULL,
  clone_graph = TRUE
)

Arguments

dataset

(torch::dataset)
The torch dataset. It should return a named list() of torch_tensor objects.

dataset_shapes

(named list() of (integer() or NULL))
The shapes of the output. Names are the elements of the list returned by the dataset. If the shape is not NULL (unknown, e.g. for images of different sizes) the first dimension must be NA to indicate the batch dimension.

graph

(Graph)
The preprocessing graph. If left NULL, no preprocessing is applied to the data and input_map, pointer, pointer_shape, and pointer_shape_predict are inferred in case the dataset returns only one element.

input_map

(character())
Character vector that must have the same length as the input of the graph. Specifies how the data from the dataset is fed into the preprocessing graph.

pointer

(character(2) | NULL)
Points to an output channel within graph: Element 1 is the PipeOp's id and element 2 is that PipeOp's output channel.

pointer_shape

(integer() | NULL)
Shape of the output indicated by pointer.

pointer_shape_predict

(integer() or NULL)
Internal use only. Used in a Graph to anticipate possible mismatches between train and predict shapes.

clone_graph

(logical(1))
Whether to clone the preprocessing graph.

Method print()

Prints the object

Usage

DataDescriptor$print(...)

Arguments

...

(any)
Unused

Method clone()

The objects of this class are cloneable with this method.

Usage

DataDescriptor$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

# Create a dataset
ds = dataset(
  initialize = function() self$x = torch_randn(10, 3, 3),
  .getitem = function(i) list(x = self$x[i, ]),
  .length = function() nrow(self$x)
)()
dd = DataDescriptor$new(ds, list(x = c(NA, 3, 3)))
dd
#> <DataDescriptor: 1 ops>
#> * dataset_shapes: [x: (NA,3,3)]
#> * input_map: (x) -> Graph
#> * pointer: nop.e90e82.x.output
#> * shape: [(NA,3,3)]
# is the same as using the converter:
as_data_descriptor(ds, list(x = c(NA, 3, 3)))
#> <DataDescriptor: 1 ops>
#> * dataset_shapes: [x: (NA,3,3)]
#> * input_map: (x) -> Graph
#> * pointer: nop.e90e82.x.output
#> * shape: [(NA,3,3)]