State-of-the-Art Voice UI Performance on Device Post date February 7, 2021 ← Quantization for Efficient Inference in Edge Devices → Towards Further Compression of Low-Bitwidth DNNs with Permuted Diagonal Matrices