Skip to main content

nll_loss_backward

teeny_kernels::nn::loss::nll

Function nll_loss_backward 

Source 
pub fn nll_loss_backward<T: Triton>(
    dy_ptr: T::Pointer<f32>,
    targets_ptr: T::Pointer<i32>,
    dx_ptr: T::Pointer<f32>,
    _n_rows: i32,
    n_cols: i32,
)
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>>> + AddOffsets<i32, 1, T::Tensor<i32>, Output = T::Tensor<T::Pointer<f32>>>,
    T::Pointer<i32>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<i32>>>,
    T::Tensor<i32>: Tensor<i32, 1>,
Expand description

NLL loss backward: dx[n, target[n]] = -dy[n], zero elsewhere.

Grid: [n_rows, 1, 1] — one CTA per batch element. The dx buffer must be zero-initialised before launch.