Skip to main content

CumulativeOps

numr::ops::traits

Trait CumulativeOps 

Source 
pub trait CumulativeOps<R: Runtime> {
    // Provided methods
    fn cumsum(&self, a: &Tensor<R>, dim: isize) -> Result<Tensor<R>> { ... }
    fn cumprod(&self, a: &Tensor<R>, dim: isize) -> Result<Tensor<R>> { ... }
    fn logsumexp(
        &self,
        a: &Tensor<R>,
        dims: &[usize],
        keepdim: bool,
    ) -> Result<Tensor<R>> { ... }
}
Expand description

Cumulative operations

Provided Methods§

Source

fn cumsum(&self, a: &Tensor<R>, dim: isize) -> Result<Tensor<R>>

Cumulative sum along a dimension

Returns the cumulative sum of elements along the specified dimension. For input [a, b, c, d], output is [a, a+b, a+b+c, a+b+c+d].

§Arguments
  • a - Input tensor
  • dim - Dimension along which to compute cumulative sum (supports negative indexing)
§Returns

Tensor with same shape as input containing cumulative sums

§Example
let a = Tensor::<CpuRuntime>::from_slice(&[1.0, 2.0, 3.0, 4.0], &[4], &device);
let result = client.cumsum(&a, 0)?; // [1, 3, 6, 10]
Source

fn cumprod(&self, a: &Tensor<R>, dim: isize) -> Result<Tensor<R>>

Cumulative product along a dimension

Returns the cumulative product of elements along the specified dimension. For input [a, b, c, d], output is [a, ab, abc, abcd].

§Arguments
  • a - Input tensor
  • dim - Dimension along which to compute cumulative product (supports negative indexing)
§Returns

Tensor with same shape as input containing cumulative products

§Example
let a = Tensor::<CpuRuntime>::from_slice(&[1.0, 2.0, 3.0, 4.0], &[4], &device);
let result = client.cumprod(&a, 0)?; // [1, 2, 6, 24]
Source

fn logsumexp( &self, a: &Tensor<R>, dims: &[usize], keepdim: bool, ) -> Result<Tensor<R>>

Log-sum-exp along specified dimensions (numerically stable)

Computes log(sum(exp(x))) in a numerically stable way: logsumexp(x) = max(x) + log(sum(exp(x - max(x))))

This is commonly used in softmax computation and log-probability calculations.

§Arguments
  • a - Input tensor
  • dims - Dimensions to reduce over
  • keepdim - If true, reduced dimensions are kept with size 1
§Returns

Tensor containing log-sum-exp values

§Example
let a = Tensor::<CpuRuntime>::from_slice(&[1.0, 2.0, 3.0], &[3], &device);
let result = client.logsumexp(&a, &[0], false)?;
// result ≈ log(exp(1) + exp(2) + exp(3)) ≈ 3.4076

Implementors§

Source§

impl CumulativeOps<CpuRuntime> for CpuClient

CumulativeOps implementation for CPU runtime.