Struct TensorView

pub struct TensorView<'a> { /* private fields */ }

A view into the original Tensor. A view borrows part of (or even the entirety of) the original Tensor data. It can NOT modify the data itself. It is normally produced by indexing or slicing a Tensor.

It can be useful for creating a new Tensor from part of another one:

Examples

use tensor_compute::{Tensor, TensorView, s};
let tensor = Tensor::from_data_and_shape(vec![1., 2., 3., 4.], vec![2, 2]);
let first_row: TensorView = tensor.slice(s![1]);
assert_eq!(first_row.shape(), &[1, 2]);
assert_eq!(first_row.to_cpu().as_contiguous_vec(), &[3., 4.]);
let new_tensor = first_row.make_contiguous(); // creates a new owned tensor from the view
assert_eq!(new_tensor.shape(), &[1, 2]);
assert_eq!(new_tensor.to_cpu().as_contiguous_vec(), &[3., 4.]);

Implementations

impl<'a> TensorView<'a>

pub async fn make_contiguous_async(&self) -> Tensor

Same as TensorView::make_contiguous, but async.

pub fn make_contiguous(&self) -> Tensor

Copies the TensorView into a standalone contiguous Tensor

Examples

use tensor_compute::{Tensor, TensorView, s};
let tensor = Tensor::from_data_and_shape(vec![1., 2., 3., 4., 5., 6., 7., 8.], vec![2, 2, 2]);
let first_row: TensorView = tensor.slice(s![1]);
assert_eq!(first_row.shape(), &[1, 2, 2]);
assert_eq!(first_row.to_cpu().as_contiguous_vec(), &[5., 6., 7., 8.]);
let new_tensor = first_row.make_contiguous(); // creates a new owned tensor from the view
assert_eq!(new_tensor.shape(), &[1, 2, 2]);
assert_eq!(new_tensor.to_cpu().as_contiguous_vec(), &[5., 6., 7., 8.]);
assert!(new_tensor.slice(s![..]).compare(&first_row));

pub async fn compare_async(&self, other: &Self) -> bool

Same as Tensor::compare_async

pub fn compare(&self, other: &Self) -> bool

Same as Tensor::compare

pub fn shape(&self) -> &VecDeque<usize>

Same as Tensor::shape

pub fn strides(&self) -> &VecDeque<usize>

Same as Tensor::strides

pub async fn to_cpu_async(&self) -> CpuTensor

Same as Tensor::to_cpu_async

pub fn to_cpu(&self) -> CpuTensor

Same as Tensor::to_cpu

Trait Implementations

impl<'a> Debug for TensorView<'a>

Beware, printing the TensorView forces a copy from GPU memory to CPU memory.

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.