Comment on page
tensor.flatten
fn flatten(self: @Tensor<T>, axis: usize) -> Tensor<T>;
Flattens the input tensor into a 2D tensor. If input tensor has shape (1, 2, 3,...n) then the output will have shape (1 * 2 * 3 * ... (axis-1), axis * (axis+1) * ... n).
self(@Tensor<T>) - The input tensor.axis(usize) - Indicate up to which input dimensions (exclusive) should be flattened.
- Panics if axis is not in the range of the input tensor's dimensions.
A new
Tensor<T> instance containing the flattened version of the input tensor.Case 1: flatten with axis 0
fn flatten_example() -> Tensor<u32> {
let tensor = TensorTrait::<u32>::new(
shape: array![2, 2, 2].span(),
data: array![0, 1, 2, 3, 4, 5, 6, 7].span(),
extra: Option::None(())
);
return tensor.flatten(0); // equivalent to tensor.reshape(1,8)
}
>>> [[0,1,2,5,4,9,6,13]]
Case 2: flatten with axis 1
use array::{ArrayTrait, SpanTrait};
use orion::operators::tensor::{TensorTrait, Tensor, U32Tensor};
fn flatten_example() -> Tensor<u32> {
let tensor = TensorTrait::<u32>::new(
shape: array![2, 2, 2].span(), data: array![0, 1, 2, 3, 4, 5, 6, 7].span(),
);
return tensor.flatten(1); // equivalent to tensor.reshape(2,4)
}
>>> [[0,1,2,3],[4,5,6,7]]
Case 3: flatten with axis 2
use array::{ArrayTrait, SpanTrait};
use orion::operators::tensor::{TensorTrait, Tensor, U32Tensor};
fn flatten_example() -> Tensor<u32> {
let tensor = TensorTrait::<u32>::new(
shape: array![2, 2, 2].span(), data: array![0, 1, 2, 3, 4, 5, 6, 7].span(),
);
return tensor.flatten(2); // equivalent to tensor.reshape(4,2)
}
>>> [[0,1],[2,3],[4,5],[6,7]]
Last modified 2mo ago