xma.functional.fused_residual_add_rmsnorm.triton_implementation.backward

fused_residual_add_rmsnorm_backward_triton_kernel(xr_ptr, xr_stride, W_ptr, W_stride, dy_ptr, dy_stride, dxr_ptr, dxr_stride, dx_ptr, dx_stride, dr_ptr, dr_stride, dW_ptr, dW_stride, s_ptr, s_stride, eps, multiplier, B, H, ATOMIC_ADD: triton.language.constexpr, BLOCK_SIZE_B: triton.language.constexpr, BLOCK_SIZE_H: triton.language.constexpr)