Huang and colleagues: GPipe micro-batch pipeline parallelism

What changed

Split training batches into micro-batches and pipelined layer partitions over accelerators to train larger sequential networks.

Scientific scope

Split training batches into micro-batches and pipelined layer partitions over accelerators to train larger sequential networks. First preprint appeared in 2018; NeurIPS publication was 2019. Pipeline bubbles and partition balance remain; source performance is not universally linear scaling.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. First preprint appeared in 2018; NeurIPS publication was 2019. Pipeline bubbles and partition balance remain; source performance is not universally linear scaling.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism
• Huang and colleagues: GPipe micro-batch pipeline parallelism — institutional research record

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction