pub unsafe extern "C" fn atg_batch_norm_backward_elemt_out(
    out__: *mut *mut C_tensor,
    out_: *mut C_tensor,
    grad_out_: *mut C_tensor,
    input_: *mut C_tensor,
    mean_: *mut C_tensor,
    invstd_: *mut C_tensor,
    weight_: *mut C_tensor,
    sum_dy_: *mut C_tensor,
    sum_dy_xmu_: *mut C_tensor,
    count_: *mut C_tensor
)