Fix the logsumexp test. (#1426)

This commit is contained in:
Laurent Mazare
2023-12-12 10:56:11 -06:00
committed by GitHub
parent 77252ffb82
commit 4cb443d00a

View File

@ -1,4 +1,4 @@
use candle_core::{test_device, test_utils, D, DType, Device, IndexOp, Result, Tensor};
use candle_core::{test_device, test_utils, DType, Device, IndexOp, Result, Tensor, D};
fn zeros(device: &Device) -> Result<()> {
let tensor = Tensor::zeros((5, 2), DType::F32, device)?;
@ -1224,25 +1224,23 @@ fn cumsum() -> Result<()> {
/// A helper function for floating point comparison. Both a and b must be 1D Tensor and contains the same amount of data.
/// Assertion passes if the difference of all pairs of a and b is smaller than epsilon.
fn assert_close(a: &Tensor, b: &Tensor, epsilon: f64) {
let a_vec: Vec<f64> = a.to_vec1().unwrap();
let b_vec: Vec<f64> = b.to_vec1().unwrap();
fn assert_close(a: &Tensor, b: &Tensor, epsilon: f64) -> Result<()> {
let a_vec: Vec<f64> = a.to_vec1()?;
let b_vec: Vec<f64> = b.to_vec1()?;
assert_eq!(a_vec.len(), b_vec.len());
for (a, b) in a_vec.iter().zip(b_vec.iter()) {
assert!((a - b).abs() < epsilon);
}
Ok(())
}
#[test]
fn logsumexp() -> Result<()> {
let input = Tensor::new(&[[1f32, 2., 3.], [4., 5., 6.]], &Device::Cpu)?;
let input = Tensor::new(&[[1f64, 2., 3.], [4., 5., 6.]], &Device::Cpu)?;
let output = input.logsumexp(D::Minus1)?;
// Expectation get from pytorch.
// The expectations obtained from pytorch.
let expected = Tensor::new(&[3.4076, 6.4076], &Device::Cpu)?;
assert_close(&output, &expected, 0.00001);
assert_close(&output, &expected, 0.00001)?;
Ok(())
}
}