pub fn gaussian_nll_loss_forward<T: Triton, const BLOCK_SIZE: i32>(
input_ptr: T::Pointer<f32>,
target_ptr: T::Pointer<f32>,
var_ptr: T::Pointer<f32>,
out_ptr: T::Pointer<f32>,
n_elements: i32,
eps_var: f32,
)where
T::I32Tensor: Tensor<i32, 1> + Comparison<i32, BoolTensor = T::BoolTensor>,
T::Pointer<f32>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<f32>>>,Expand description
Element-wise Gaussian NLL loss forward.
out = 0.5 * (log(var) + (input - target)^2 / var)var is clamped to eps_var from below for numerical stability.