PacTrain combines model pruning, gradient sparsity enforcement, and ternary quantization to make gradient synchronization all-reduce compatible and communication-efficient.
Lamda: Language models for dialog applications,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
PacTrain: Pruning and Adaptive Sparse Gradient Compression for Efficient Collective Communication in Distributed Deep Learning
PacTrain combines model pruning, gradient sparsity enforcement, and ternary quantization to make gradient synchronization all-reduce compatible and communication-efficient.