Cambricon-Q: A Hybrid Architecture for Efficient Training
ISCA '21 Proceedings of the 48th Annual International Symposium on Computer Architecture（2021）
关键词
hybrid architecture,neural network training,computation costs,negligible training accuracy loss,DNN accelerators,customized architecture,efficient quantized training,negligible accuracy loss,ASIC acceleration core,near-data-processing engine,statistic-based quantization,specialized computing units,statistical analysis,NDP engine,high-precision weights,energy efficiency,DNN training,accuracy degradation,precision training,quantized algorithms
AI 理解论文
溯源树
样例
生成溯源树，研究论文发展脉络
Chat Paper
正在生成论文摘要