Microsoft’s Open Sourced a New Library for Extreme Compression of Deep Learning Models

DeepSpeed Compression combines compression and system optimization techniques for building smaller and more efficient deep learning architectures.