The paper proposes a novel analysis which gives a dimension free bound for Low precision training of ML models. The bound is interesting and can have good impact on the field