Comment on page
tensor.onehot
fn onehot(self: @Tensor<T>, depth: usize, axis: Option<usize>, values: Span<usize>) -> Tensor<usize>;
Produces one-hot tensor based on input.
self(@Tensor<T>) - The input tensor.depth(usize) - Scalar or Rank 1 tensor containing exactly one element, specifying the number of classes in one-hot tensor.axis(Option<bool>) - Axis along which one-hot representation in added. Default: axis=-1.values(Span<usize>) - Rank 1 tensor containing exactly two elements, in the format [off_value, on_value]
- Panics if values is not equal to 2.
A new
Tensor<T> one-hot encode of the input tensor.Constrain input and output types to fixed point tensors.
use array::{ArrayTrait, SpanTrait};
use orion::operators::tensor::{TensorTrait, Tensor, FP8x23Tensor};
use orion::numbers::{FP8x23, FixedTrait};
fn onehot_example() -> Tensor<FP8x23> {
let tensor = TensorTrait::<FP8x23>::new(
shape: array![2,2].span(),
data: array![
FixedTrait::new_unscaled(0, false),
FixedTrait::new_unscaled(1, false),
FixedTrait::new_unscaled(2, false),
FixedTrait::new_unscaled(3, false),
]
.span(),
);
return tensor.onehot(depth: 3, axis: Option::None(()), values: array![0, 1].span());
}
>>> [[1. 0. 0.]
[0. 1. 0.]
[0. 0. 1.]]
Last modified 2mo ago