Skip to main content

yolo_ciou_loss_forward

vision_rs::models::yolo::kernels::loss::ciou

Function yolo_ciou_loss_forward 

Source 
pub fn yolo_ciou_loss_forward<T: Triton, D: Float, const BLOCK_N: i32>(
    pred_ptr: T::Pointer<D>,
    target_ptr: T::Pointer<D>,
    loss_ptr: T::Pointer<D>,
    iou_ptr: T::Pointer<D>,
    v_ptr: T::Pointer<D>,
    alpha_ptr: T::Pointer<D>,
    N: i32,
)
where
    T::I32Tensor: Tensor<i32, 1> + Comparison<i32, BoolTensor = T::BoolTensor>,
    T::Pointer<D>: AddOffsets<i32, 1, T::I32Tensor, Output = T::Tensor<T::Pointer<D>>>,
Expand description

Fused CIoU loss forward: predicted and target XYWH boxes → per-anchor loss plus saved activations (iou, v, alpha) needed by the backward pass.

Grid: cdiv(N, BLOCK_N) — one CTA per anchor tile.