Towards Cheaper Inference in Deep Networks with Lower Bit-Width Accumulators
Keywords: Quantization, Inference, Accumulation, Efficient Neural Networks, Quantized Neural Networks
TL;DR: We show that Neural networks can be fine tuned to work with 12-bit, floating point accumulators, and we propose new methods to reduce the accumulator bit-width further
Abstract: The majority of the research on the quantization of Deep Neural Networks (DNNs) is focused on reducing the precision of tensors visible by high-level frameworks (e.g., weights, activations, and gradients). However, current hardware still relies on high-accuracy core operations. Most significant is the operation of accumulating products. This high-precision accumulation operation is gradually becoming the main computational bottleneck. This is because, so far, the usage of low-precision accumulators led to a significant degradation in performance. In this work, we present a simple method to train and fine-tune high-end DNNs, to allow, for the first time, utilization of cheaper, $12$-bits accumulators, with no significant degradation in accuracy. Lastly, we show that as we decrease the accumulation precision further, using fine-grained gradient approximations can improve the DNN accuracy.
Submission Number: 23