wrapping a keras optimizer to implement gradient accumulation - View it on GitHub