SWALP Stochastic Weight Averaging in Low Precision Training

SWALP:StochasticWeightAveraginginLow-PrecisionTrainingGuandaoYang1TianyiZhang1PolinaKirichenko1JunwenBai1AndrewGordonWilson1ChristopherDeSa1Abstractandaccumulategradientinformationinhigherprecision...