Enregistré dans:
Détails bibliographiques
Auteurs principaux: Subramanian, Shashank, Rrapaj, Ermal, Harrington, Peter, Chheda, Smeet, Farrell, Steven, Austin, Brian, Williams, Samuel, Wright, Nicholas, Bhimji, Wahid
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2410.00273
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866914961252417536
author Subramanian, Shashank
Rrapaj, Ermal
Harrington, Peter
Chheda, Smeet
Farrell, Steven
Austin, Brian
Williams, Samuel
Wright, Nicholas
Bhimji, Wahid
author_facet Subramanian, Shashank
Rrapaj, Ermal
Harrington, Peter
Chheda, Smeet
Farrell, Steven
Austin, Brian
Williams, Samuel
Wright, Nicholas
Bhimji, Wahid
contents Generative AI, in particular large transformer models, are increasingly driving HPC system design in science and industry. We analyze performance characteristics of such transformer models and discuss their sensitivity to the transformer type, parallelization strategy, and HPC system features (accelerators and interconnects). We utilize a performance model that allows us to explore this complex design space and highlight its key components. We find that different transformer types demand different parallelism and system characteristics at different training regimes. Large Language Models are performant with 3D parallelism and amplify network needs only at pre-training scales with reduced dependence on accelerator capacity and bandwidth. On the other hand, long-sequence transformers, representative of scientific foundation models, place a more uniform dependence on network and capacity with necessary 4D parallelism. Our analysis emphasizes the need for closer performance modeling of different transformer types keeping system features in mind and demonstrates a path towards this. Our code is available as open-source.
format Preprint
id arxiv_https___arxiv_org_abs_2410_00273
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Comprehensive Performance Modeling and System Design Insights for Foundation Models
Subramanian, Shashank
Rrapaj, Ermal
Harrington, Peter
Chheda, Smeet
Farrell, Steven
Austin, Brian
Williams, Samuel
Wright, Nicholas
Bhimji, Wahid
Machine Learning
Distributed, Parallel, and Cluster Computing
Generative AI, in particular large transformer models, are increasingly driving HPC system design in science and industry. We analyze performance characteristics of such transformer models and discuss their sensitivity to the transformer type, parallelization strategy, and HPC system features (accelerators and interconnects). We utilize a performance model that allows us to explore this complex design space and highlight its key components. We find that different transformer types demand different parallelism and system characteristics at different training regimes. Large Language Models are performant with 3D parallelism and amplify network needs only at pre-training scales with reduced dependence on accelerator capacity and bandwidth. On the other hand, long-sequence transformers, representative of scientific foundation models, place a more uniform dependence on network and capacity with necessary 4D parallelism. Our analysis emphasizes the need for closer performance modeling of different transformer types keeping system features in mind and demonstrates a path towards this. Our code is available as open-source.
title Comprehensive Performance Modeling and System Design Insights for Foundation Models
topic Machine Learning
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2410.00273