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.

Sign up for GitHub

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

Add cumulative sum tensor operation #1722

Closed
wants to merge 11 commits into from
Closed

Add cumulative sum tensor operation #1722

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
b7188aa
Begin draft of cumulative sum: interfaces
allenqm May 3, 2024
c51d7be
format
allenqm May 7, 2024
729778e
add torch implementation
allenqm May 7, 2024
ed7681e
add candle implementation
allenqm May 7, 2024
9175f92
add ndarray implementation
allenqm May 7, 2024
3a477a2
add autodiff implementation
allenqm May 7, 2024
46cc2c8
add tensor tests
allenqm May 7, 2024
91c27b4
remove int overflow test
allenqm May 7, 2024
fdc5c95
add cumsum source to burn-import. tests/ in subsequent commits.
allenqm May 7, 2024
8b9b39e
add burn import tests
allenqm May 8, 2024
dab2905
rename cumsum_dim to cumsum
allenqm May 8, 2024
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
13 changes: 12 additions & 1 deletion crates/burn-ndarray/src/ops/int_tensor.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@ use burn_tensor::{Distribution, Reader};

use burn_tensor::ElementConversion;
use core::ops::Range;
use ndarray::IntoDimension;
use ndarray::{Axis, IntoDimension};

// Current crate
use crate::element::ExpElement;
Expand Down Expand Up @@ -286,6 +286,17 @@ impl<E: FloatNdArrayElement> IntTensorOps<Self> for NdArray<E> {
NdArrayMathOps::sum_dim(tensor, dim)
}

fn int_cumsum_dim<const D: usize>(
tensor: NdArrayTensor<i64, D>,
dim: usize,
) -> NdArrayTensor<i64, D> {
let mut array = tensor.array.clone().into_owned();
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See comment for float_cumsum


array.accumulate_axis_inplace(Axis(dim), |&prev, curr| *curr += prev);

NdArrayTensor::new(array.to_shared())
}

fn int_prod<const D: usize>(tensor: NdArrayTensor<i64, D>) -> NdArrayTensor<i64, 1> {
NdArrayMathOps::prod(tensor)
}
Expand Down
13 changes: 12 additions & 1 deletion crates/burn-ndarray/src/ops/tensor.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
// Language
use alloc::vec::Vec;
use core::ops::Range;
use ndarray::IntoDimension;
use ndarray::{Axis, IntoDimension};

// Current crate
use super::{matmul::matmul, NdArrayMathOps, NdArrayOps};
Expand Down Expand Up @@ -338,6 +338,17 @@ impl<E: FloatNdArrayElement> FloatTensorOps<Self> for NdArray<E> {
NdArrayMathOps::sum_dim(tensor, dim)
}

fn float_cumsum_dim<const D: usize>(
tensor: NdArrayTensor<E, D>,
dim: usize,
) -> NdArrayTensor<E, D> {
let mut array = tensor.array.clone().into_owned();
Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I believe the underlying array struct of tensor needs to be cloned, since NdArray's method for accumulating elements along an axis modifies an array's data inplace. Referring to this method

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well float_cumsum takes ownership of the tensor, so I don't think the clone is required here.


array.accumulate_axis_inplace(Axis(dim), |&prev, curr| *curr += prev);

NdArrayTensor::new(array.to_shared())
}

fn float_argmax<const D: usize>(
tensor: NdArrayTensor<E, D>,
dim: usize,
Expand Down