Output Shape of a Reshape

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