Struct GpuTensor 

pub struct GpuTensor {
    pub buffer: Buffer,
    pub shape: Vec<usize>,
    pub dtype: DType,
    pub numel: usize,
    pub device: Arc<GpuDevice>,
}

A tensor stored on the GPU as an f32 buffer.

Fields

buffer: Buffer

shape: Vec<usize>

dtype: DType

numel: usize

device: Arc<GpuDevice>

Implementations

impl GpuTensor

pub fn from_cpu(tensor: &TlTensor, device: Arc<GpuDevice>) -> Self

Upload a CPU TlTensor (f64) to GPU as f32.

pub fn from_f32(data: &[f32], shape: Vec<usize>, device: Arc<GpuDevice>) -> Self

Create a GpuTensor from f32 data.

pub fn to_cpu(&self) -> Result<TlTensor, String>

Download GPU tensor to CPU as TlTensor (f64).

pub fn read_f32(&self) -> Result<Vec<f32>, String>

Read raw f32 data from the GPU buffer.

pub fn byte_size(&self) -> u64

Get the total byte size of the buffer.

Trait Implementations

impl Clone for GpuTensor

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for GpuTensor

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

Formats the value using the given formatter.

impl Display for GpuTensor

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

Formats the value using the given formatter.