pub fn channel_bias_add_backward<T: Triton, D: Float, const BLOCK_N: i32>(
dy_ptr: T::Pointer<D>,
dx_ptr: T::Pointer<D>,
dbias_ptr: T::Pointer<D>,
N: i32,
C: 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
Backward pass for channel bias add.
dx = dy (identity), dbias[c] = sum over N of dy[n, c].
Grid: [C] — one CTA per channel.