Data movement limits to frontier model training
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913577110077440 |
|---|---|
| author | Erdil, Ege Schneider-Joseph, David |
| author_facet | Erdil, Ege Schneider-Joseph, David |
| contents | We present a theoretical model of distributed training, and use it to analyze how far dense and sparse training runs can be scaled. Under our baseline assumptions, given a three month training duration, data movement bottlenecks begin to significantly lower hardware utilization for training runs exceeding about $10^{28}$ FLOP, two orders of magnitude above the largest training run to date, suggesting the arrival of fundamental barriers to scaling in three years given recent rates of growth. A training run exceeding about $10^{31}$ FLOP is infeasible even at low utilization. However, more aggressive batch size scaling and/or shorter and fatter model shapes, if achievable, have the potential to permit much larger training runs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2411_01137 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Data movement limits to frontier model training Erdil, Ege Schneider-Joseph, David Distributed, Parallel, and Cluster Computing Artificial Intelligence Machine Learning We present a theoretical model of distributed training, and use it to analyze how far dense and sparse training runs can be scaled. Under our baseline assumptions, given a three month training duration, data movement bottlenecks begin to significantly lower hardware utilization for training runs exceeding about $10^{28}$ FLOP, two orders of magnitude above the largest training run to date, suggesting the arrival of fundamental barriers to scaling in three years given recent rates of growth. A training run exceeding about $10^{31}$ FLOP is infeasible even at low utilization. However, more aggressive batch size scaling and/or shorter and fatter model shapes, if achievable, have the potential to permit much larger training runs. |
| title | Data movement limits to frontier model training |
| topic | Distributed, Parallel, and Cluster Computing Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2411.01137 |