huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 18:48:51 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use a reference for the device.

Browse Source
This commit is contained in:
laurent
2023-06-21 19:55:57 +01:00
parent deb6091099
commit fcb4e6b84f
5 changed files with 23 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/basics.rs
View File

@ -2,7 +2,7 @@ use anyhow::Result;
use candle::{Device, Tensor};
fn main() -> Result<()> {
let x = Tensor::var(&[3f32, 1., 4.], Device::Cpu)?;
let x = Tensor::var(&[3f32, 1., 4.], &Device::Cpu)?;
let y = (((&x * &x)? + &x * 5f64)? + 4f64)?;
println!("{:?}", y.to_vec1::<f32>()?);
Ok(())

7
examples/cuda_basics.rs
View File

@ -1,9 +1,12 @@
use anyhow::Result;
use candle::{DType, Device, Tensor};
use candle::{Device, Tensor};
fn main() -> Result<()> {
let device = Device::new_cuda(0)?;
let x = Tensor::zeros(4, DType::F32, device)?;
let x = Tensor::new(&[3f32, 1., 4., 1., 5.], &device)?;
let y = Tensor::new(&[2f32, 7., 1., 8., 2.], &device)?;
println!("{:?}", x.to_vec1::<f32>()?);
let z = (x + y)?;
println!("{:?}", z.to_vec1::<f32>()?);
Ok(())
}

22
src/tensor.rs
View File

@ -84,7 +84,7 @@ impl Tensor {
fn ones_impl<S: Into<Shape>>(
shape: S,
dtype: DType,
device: Device,
device: &Device,
is_variable: bool,
) -> Result<Self> {
let shape = shape.into();
@ -101,22 +101,22 @@ impl Tensor {
Ok(Self(Arc::new(tensor_)))
}
pub fn ones<S: Into<Shape>>(shape: S, dtype: DType, device: Device) -> Result<Self> {
pub fn ones<S: Into<Shape>>(shape: S, dtype: DType, device: &Device) -> Result<Self> {
Self::ones_impl(shape, dtype, device, false)
}
pub fn ones_var<S: Into<Shape>>(shape: S, dtype: DType, device: Device) -> Result<Self> {
pub fn ones_var<S: Into<Shape>>(shape: S, dtype: DType, device: &Device) -> Result<Self> {
Self::ones_impl(shape, dtype, device, true)
}
pub fn ones_like(&self) -> Result<Self> {
Tensor::ones(self.shape(), self.dtype(), self.device())
Tensor::ones(self.shape(), self.dtype(), &self.device())
}
fn zeros_impl<S: Into<Shape>>(
shape: S,
dtype: DType,
device: Device,
device: &Device,
is_variable: bool,
) -> Result<Self> {
let shape = shape.into();
@ -133,21 +133,21 @@ impl Tensor {
Ok(Self(Arc::new(tensor_)))
}
pub fn zeros<S: Into<Shape>>(shape: S, dtype: DType, device: Device) -> Result<Self> {
pub fn zeros<S: Into<Shape>>(shape: S, dtype: DType, device: &Device) -> Result<Self> {
Self::zeros_impl(shape, dtype, device, false)
}
pub fn zeros_var<S: Into<Shape>>(shape: S, dtype: DType, device: Device) -> Result<Self> {
pub fn zeros_var<S: Into<Shape>>(shape: S, dtype: DType, device: &Device) -> Result<Self> {
Self::zeros_impl(shape, dtype, device, true)
}
pub fn zeros_like(&self) -> Result<Self> {
Tensor::zeros(self.shape(), self.dtype(), self.device())
Tensor::zeros(self.shape(), self.dtype(), &self.device())
}
pub fn new_impl<A: crate::device::NdArray>(
array: A,
device: Device,
device: &Device,
is_variable: bool,
) -> Result<Self> {
let shape = array.shape()?;
@ -164,11 +164,11 @@ impl Tensor {
Ok(Self(Arc::new(tensor_)))
}
pub fn new<A: crate::device::NdArray>(array: A, device: Device) -> Result<Self> {
pub fn new<A: crate::device::NdArray>(array: A, device: &Device) -> Result<Self> {
Self::new_impl(array, device, false)
}
pub fn var<A: crate::device::NdArray>(array: A, device: Device) -> Result<Self> {
pub fn var<A: crate::device::NdArray>(array: A, device: &Device) -> Result<Self> {
Self::new_impl(array, device, true)
}

2
tests/grad_tests.rs
View File

@ -3,7 +3,7 @@ use candle::{Device, Tensor};
#[test]
fn simple_grad() -> Result<()> {
let x = Tensor::var(&[3f32, 1., 4.], Device::Cpu)?;
let x = Tensor::var(&[3f32, 1., 4.], &Device::Cpu)?;
let y = (((&x * &x)? + &x * 5f64)? + 4f64)?;
let grads = y.backward()?;
let grad_x = grads.get(&x).context("no grad for x")?;

10
tests/tensor_tests.rs
View File

@ -2,7 +2,7 @@ use candle::{DType, Device, Result, Tensor};
#[test]
fn zeros() -> Result<()> {
let tensor = Tensor::zeros((5, 2), DType::F32, Device::Cpu)?;
let tensor = Tensor::zeros((5, 2), DType::F32, &Device::Cpu)?;
let (dim1, dim2) = tensor.shape().r2()?;
assert_eq!(dim1, 5);
assert_eq!(dim2, 2);
@ -11,7 +11,7 @@ fn zeros() -> Result<()> {
#[test]
fn add_mul() -> Result<()> {
let tensor = Tensor::new(&[3f32, 1., 4.], Device::Cpu)?;
let tensor = Tensor::new(&[3f32, 1., 4.], &Device::Cpu)?;
let dim1 = tensor.shape().r1()?;
assert_eq!(dim1, 3);
let content: Vec<f32> = tensor.to_vec1()?;
@ -28,7 +28,7 @@ fn add_mul() -> Result<()> {
#[test]
fn tensor_2d() -> Result<()> {
let data = &[[3f32, 1., 4., 1., 5.], [2., 1., 7., 8., 2.]];
let tensor = Tensor::new(data, Device::Cpu)?;
let tensor = Tensor::new(data, &Device::Cpu)?;
let dims = tensor.shape().r2()?;
assert_eq!(dims, (2, 5));
let content: Vec<Vec<f32>> = tensor.to_vec2()?;
@ -39,9 +39,9 @@ fn tensor_2d() -> Result<()> {
#[test]
fn binary_op() -> Result<()> {
let data = &[[3f32, 1., 4., 1., 5.], [2., 1., 7., 8., 2.]];
let tensor = Tensor::new(data, Device::Cpu)?;
let tensor = Tensor::new(data, &Device::Cpu)?;
let data2 = &[[5f32, 5., 5., 5., 5.], [2., 1., 7., 8., 2.]];
let tensor2 = Tensor::new(data2, Device::Cpu)?;
let tensor2 = Tensor::new(data2, &Device::Cpu)?;
let tensor = (&tensor + (&tensor * &tensor)? / (&tensor + &tensor2))?;
let dims = tensor.shape().r2()?;
assert_eq!(dims, (2, 5));