• 2025-01-05Training Deep Neural Networks with 8-bit Floating Point Numbers
    目录概主要内容WangN.,ChoiJ.,BrandD.,ChenC.andGopalakrishnanK.Trainingdeepneuralnetworkswith8-bitfloatingpointnumbers.NeurIPS,2018.概本文提出了一种8-bit的训练方式.主要内容本文想要实现8-bit的训练,作者认为主要挑战是两个向量的