Inference of Long-Short Term Memory networks at software-equivalent accuracy using 2.5M analog Phase Change Memory devices

Abstract

We report accuracy for forward inference of long-short-term-memory (LSTM) networks using weights programmed into the conductances of $> 2.5\text{M}$ phase-change memory (PCM) devices. We demonstrate strategies for software weight-mapping and programming of hardware analog conductances that provide accurate weight programming despite significant device variability. Inference accuracy very close to software-model baselines is achieved on several language modeling tasks.