Q-BERT: Hessian Based Ultra Low Precision Quantization of BERT---Q-BERT:基于Hessian的超低精度BERT量化
HessianBasedUltraLowPrecisionQuantizationofBERT---Q-BERT:基于Hessian的超低精度BERT量化Abstract1RelatedWorkModelcompression模型压缩CompressedNLPmodel2Methodology2.1Quantizationprocess2.2Mixedprecisionqu