Fetching the paper…

ZeroQuant: Efficient and Affordable Post-Training Quantization for Large-Scale Transformers · Around