MPipeMoE speeds up MoE training by adaptively pipelining token batches and reusing memory buffers across partitions, achieving up to 2.8x speedup and 47% memory reduction over FasterMoE.
Bert: Pre-training of deep bidirectional transformers for language understanding,
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
background 1representative citing papers
citing papers explorer
-
MPipeMoE: Memory Efficient MoE for Pre-trained Models with Adaptive Pipeline Parallelism
MPipeMoE speeds up MoE training by adaptively pipelining token batches and reusing memory buffers across partitions, achieving up to 2.8x speedup and 47% memory reduction over FasterMoE.