Shoeybi and colleagues: Megatron-LM intra-layer model parallelism

What changed

Partitioned Transformer operations within layers across accelerators with targeted communication, enabling large language-model training.

Scientific scope

Partitioned Transformer operations within layers across accelerators with targeted communication, enabling large language-model training. One tensor-parallel training method, not a separate invention for every trained model size. Pipeline parallelism is complementary rather than identical.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. One tensor-parallel training method, not a separate invention for every trained model size. Pipeline parallelism is complementary rather than identical.

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.

• Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism

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