MixPipe: Efficient Bidirectional Pipeline Parallelism for Training Large-Scale Models.

Weigang Zhang Biyu Zhou Xuehai Tang Zhaoxing Wang Songlin Hu

Published in: DAC (2023)

Keyphrases

statistical models
neural network model
real world
probabilistic model
supervised learning
database
data sets
neural network
least squares
machine learning algorithms
experimental data
message passing
mathematical models
training algorithm
accurate models
parallel execution