to_dim

rstsr_openblas::prelude_dev

Function to_dim 

Source 
pub fn to_dim<R, T, B, D, D2>(
    tensor: &TensorBase<Storage<R, T, B>, D>,
) -> TensorBase<Storage<DataRef<'_, <B as DeviceRawAPI<T>>::Raw>, T, B>, D2>
where
    D: DimAPI + DimIntoAPI<D2>,
    D2: DimAPI,
    R: DataAPI<Data = <B as DeviceRawAPI<T>>::Raw>,
    B: DeviceAPI<T>,
Expand description

Convert layout to the other dimension.

This function is intended to be used as associated method TensorAny::into_dim. We just provide the API document here.

§Parameters

  • tensor: &TensorAny<R, T, B, D>

    • The input tensor.

  • generic parameter D2: DimAPI

    • Target dimension type.
    • IxD or IxDyn (equilvalent to Vec<usize>) can be used for dynamic dimension.
    • Ix<N> (given const N: usize, equilvalent to [usize; N]) can be used for static dimension.
    • Dynamic dimensionality is usually more preferred for general use cases, while static dimensionality is more suitable for performance-critical code, especially frequent indexing and non-contiguous memory access.

§Returns

  • TensorView<'_, T, B, D2>

    • A view of the input tensor with the layout converted to the target dimension type.
    • The underlying data is not copied; only the layout of the view is modified.
    • If you want to convert the tensor itself (taking the ownership instead of returning view), use into_dim instead.

§Examples

In most cases, RSTSR will generate dynamic dimension (IxD) in most cases, including indexing slice, reshaping reshape, creation asarray.

You can debug print tensor or it’s layout to verify the dimensionality, or call const_ndim to check whether the dimension is static or dynamic.

use rstsr::prelude::*;
let a = rt::arange(6).into_shape([2, 3]); // shape: (2, 3), IxD
println!("a: {:?}", a);
// `None` here indicates dynamic dimension
assert_eq!(a.shape().const_ndim(), None);

You can convert the dimension to static dimension by associate method:

let b = a.to_dim::<Ix2>(); // shape: (2, 3), Ix2
assert_eq!(b.shape().const_ndim(), Some(2));

You can use .to_dim::<IxD>() or .to_dyn() to convert back to dynamic dimension:

let c = b.to_dyn(); // shape: (2, 3), IxD
assert_eq!(c.shape().const_ndim(), None);

§Panics

  • The shape of the tensor is not compatible with the target dimension type.
use rstsr::prelude::*;
let a = rt::arange(6).into_shape([2, 3]); // shape: (2, 3), IxD
let b = a.to_dim::<Ix3>(); // shape: (2, 3), Ix3, panics

§See also

§Similar functions from other crates/libraries

§Variants of this function