fix hoc _update_best()

fix __ issues

rollback metric.py _record_explainer() add _

add docstring

fix docstrings newline

From: @ngtony
Reviewed-by: 
Signed-off-by:

fix pylint problem

fix review comments

fix review comments

enhance register_saliency() checking

fix error type

fix dataset class checking

fix dataset class checking

fix dataset class checking

fix fd open mode

better naming

better naming

fix review comments

fix review comments

fix review comments

fix duplicated benchmarker classes checking

fix comment format

fix comments

This commit provides APIs for user to use the widely used attribution methods to explain DL models and the evaluation methods to quantify the explanations. With combination of MindInsight, the user can have a friendly visualization on their models.