mini-ode 0.1.5

A minimalistic ODE solvers library built on top of PyTorch
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

use crate::optimizers::differentiate;
use tch::{IndexOp, Tensor};

#[test]
fn test_differentiate_case1() {
    let x = Tensor::from_slice(&[1.0f32, 2.0f32]);
    let grad = differentiate(&|y| y.i(0) + y.i(1), &x).unwrap();
    let expected = Tensor::from_slice(&[1f32, 1f32]);
    assert_eq!(grad, expected);
}

#[test]
fn test_differentiate_case2() {
    let x = Tensor::from_slice(&[1.0f32, 2.0f32]);
    let grad = differentiate(&|y| y.i(0) * y.i(0) + y.i(1) * y.i(1) * y.i(1) + y.i(1), &x).unwrap();
    let expected = Tensor::from_slice(&[2f32, 13f32]);
    assert_eq!(grad, expected);
}

#[test]
fn test_differentiate_case3() {
    let x = Tensor::from_slice(&[-1.0f32, 1.5f32]);
    let grad = differentiate(&|y| y.i(0).exp() + y.i(1), &x).unwrap();
    let expected = Tensor::stack(&[Tensor::from(-1f32).exp(), Tensor::from(1f32)], 0);
    assert_eq!(grad, expected);
}