Towards Cheaper Inference in Deep Networks with Lower Bit-Width Accumulators

Re-assign community