Function auto_diff::op::_gradient_checker

pub fn _gradient_checker(
    op: &mut dyn OpTrait, 
    one_input: &[Tensor], 
    input_mask: Option<&[bool]>, 
    step: Option<Tensor>, 
    tolerance: Option<Tensor>
) -> bool

Verify the gradient implementation is right.