reshape_output_shape: Output Shape of a Reshape

In mlr3torch: Deep Learning with 'mlr3'

Output Shape of a Reshape

Description

The shape of torch_reshape(x, shape), which is what nn_reshape() infers. The dimension that torch infers from the number of elements is resolved here whenever that number is known, and stays unknown otherwise.

Usage

reshape_output_shape(shape_in, shape, id)

Arguments

shape_in	(integer()) The input shape, with NA for the dimensions whose size is unknown.
shape	(integer() | ⁠function()⁠) The target shape, where -1 marks the dimension that torch infers from the number of elements. A ⁠function(shape)⁠ of the input shape is called on it and must return such a vector, which lets a reshape be expressed for inputs whose sizes are not known in advance, e.g. ⁠\(shape) c(shape[1:2], 10)⁠.
id	(character(1)) The id of the PipeOp, which the error messages name.

Value

(integer()) The output shape.

See Also

Other Shape Inference: infer_shapes(), shape_to_str()

Examples

reshape_output_shape(c(NA, 3, 4), c(-1, 12), id = "nn_reshape")
reshape_output_shape(c(NA, 3, 4), \(shape) c(shape[1], 12), id = "nn_reshape")

mlr3torch documentation built on Oct. 30, 2026, 5:10 p.m.