Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Zeros, ones, and full constructors #2

Merged
merged 5 commits into from
Dec 16, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
53 changes: 49 additions & 4 deletions src/tensor/constructors.rs
Original file line number Diff line number Diff line change
Expand Up @@ -31,12 +31,54 @@ impl<T: RawDataType> Tensor<T> {
ndims: D,
}
}

pub fn full(n: T, shape: Vec<usize>) -> Self {
assert!(!shape.is_empty(), "Cannot create a zero-dimension tensor!");

let vector_ns = vec![n; shape.iter().product()];

let mut stride = vec![0; shape.len()];

let ndims = shape.len();

let mut p = 1;
for i in (0..ndims).rev() {
stride[i] = p;
p *= shape[i];
}

Self {
data: DataOwned::new(vector_ns),
shape,
stride,
ndims,
}
}

pub fn zeros(shape: Vec<usize>) -> Self
where
T: RawDataType + From<bool>,
{
Self::full(false.into(), shape)
}

pub fn ones(shape: Vec<usize>) -> Self
where
T: RawDataType + From<bool>,
{
Self::full(true.into(), shape)
}
}

impl<T: RawDataType> TensorView<T> {
pub(super) fn from<B>(tensor: &TensorBase<B>, offset: usize, shape: Vec<usize>, stride: Vec<usize>) -> Self
pub(super) fn from<B>(
tensor: &TensorBase<B>,
offset: usize,
shape: Vec<usize>,
stride: Vec<usize>,
) -> Self
where
B: DataBuffer<DType=T>,
B: DataBuffer<DType = T>,
{
let ndims = shape.len();

Expand All @@ -46,9 +88,12 @@ impl<T: RawDataType> TensorView<T> {
// }
//
// the following code is equivalent to the above loop
let len = shape.iter().zip(stride.iter())
let len = shape
.iter()
.zip(stride.iter())
.map(|(&axis_length, &axis_stride)| axis_stride * (axis_length - 1))
.sum::<usize>() + 1;
.sum::<usize>()
+ 1;

let data = DataView::from_buffer(&tensor.data, offset, len);

Expand Down
2 changes: 1 addition & 1 deletion src/tensor/dtype.rs
Original file line number Diff line number Diff line change
Expand Up @@ -15,4 +15,4 @@ impl RawDataType for i128 {}
impl RawDataType for f32 {}
impl RawDataType for f64 {}

impl RawDataType for bool {}
impl RawDataType for bool {}
151 changes: 149 additions & 2 deletions tests/tensor.rs
Original file line number Diff line number Diff line change
Expand Up @@ -420,16 +420,163 @@ fn unsqueeze_random_dimension_first_axis() {

#[test]
fn unsqueeze_random_dimension_axis_1() {
let a: Tensor<i32> = Tensor::from([[1, 2, 3], [4, 5, 6]]);
let a = Tensor::from([[1, 2, 3], [4, 5, 6]]);
let b = a.unsqueeze(Axis(1));
assert_eq!(b.shape(), &[2, 1, 3]);
assert_eq!(b.stride(), &[3, 3, 1]);
}

#[test]
fn unsqueeze_random_dimension_last_axis() {
let a: Tensor<i32> = Tensor::from([[1, 2, 3], [4, 5, 6]]);
let a = Tensor::from([[1, 2, 3], [4, 5, 6]]);
let b = a.unsqueeze(Axis(2));
assert_eq!(b.shape(), &[2, 3, 1]);
assert_eq!(b.stride(), &[3, 1, 1]);
}

#[test]
fn full_i32(){
let a = Tensor::full(3, vec![2, 3]);
assert_eq!(a.shape(), &[2, 3]);
assert_eq!(a.stride(), &[3, 1]);
let b = a.flatten();
let b_len = b.len().clone();
for i in 0..b_len {
assert_eq!(b[i], 3i32);
}
}

#[test]
fn full_f64(){
let a = Tensor::full(3.2,vec![4, 6, 2]);
assert_eq!(a.shape(), &[4, 6, 2]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], 3.2f64);
}
}

#[test]
fn full_bool(){
let a: Tensor<bool> = Tensor::full(true,vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], true);
}
}

#[test]
fn ones_u8(){
let a: Tensor<u8> = Tensor::ones(vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], 1u8);
}
}

#[test]
fn ones_i32(){
let a: Tensor<i32> = Tensor::ones(vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], 1i32);
}
}

#[test]
fn ones_1d(){
let a: Tensor<u8> = Tensor::ones(vec![4]);
assert_eq!(a.shape(), &[4]);
let a_len = *a.len();
for i in 0..a_len {
assert_eq!(a[i], 1u8);
}
}

#[test]
fn ones_f64(){
let a: Tensor<f64> = Tensor::ones(vec![4]);
assert_eq!(a.shape(), &[4]);
let a_len = *a.len();
for i in 0..a_len {
assert_eq!(a[i], 1f64);
}
}

#[test]
fn ones_bool(){
let a: Tensor<bool> = Tensor::ones(vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], true);
}
}

#[test]
fn zeroes_u8(){
let a: Tensor<u8> = Tensor::zeros(vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], 0u8);
}
}

#[test]
fn zeroes_i32(){
let a: Tensor<i32> = Tensor::zeros(vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], 0i32);
}
}

#[test]
fn zeroes_1d(){
let a: Tensor<u8> = Tensor::zeros(vec![4]);
assert_eq!(a.shape(), &[4]);
let a_len = *a.len();
for i in 0..a_len {
assert_eq!(a[i], 0u8);
}
}

#[test]
fn zeroes_f64(){
let a: Tensor<f64> = Tensor::zeros(vec![4]);
assert_eq!(a.shape(), &[4]);
let a_len = *a.len();
for i in 0..a_len {
assert_eq!(a[i], 0f64);
}
}

#[test]
fn zeroes_bool(){
let a: Tensor<bool> = Tensor::zeros(vec![3, 5, 3]);
assert_eq!(a.shape(), &[3, 5, 3]);
assert_eq!(a.stride(), &[15, 3, 1]);
let b = a.flatten();
let b_len = *b.len();
for i in 0..b_len {
assert_eq!(b[i], false);
}
}
Loading