CFP: Efficient Optimization of Intra-Operator Parallelism Plans for Large Model Training
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911040601587712 |
|---|---|
| author | Hu, Weifang Shi, Xuanhua Zhang, Yunkai Wu, Chang Peng, Xuan Zhai, Jiaqi Jin, Hai Qian, Xuehai Xue, Jingling Zhou, Yongluan |
| author_facet | Hu, Weifang Shi, Xuanhua Zhang, Yunkai Wu, Chang Peng, Xuan Zhai, Jiaqi Jin, Hai Qian, Xuehai Xue, Jingling Zhou, Yongluan |
| contents | Optimizing the parallel training of large models requires exploring intra-operator parallelism plans for a computation graph that typically contains tens of thousands of primitive operators. While the optimization of parallel data processing graphs has been extensively researched in database systems, the vast search space makes it challenging to apply traditional database query optimization methods and algorithms. This paper introduces CFP, an optimization system for intra-operator parallelism that significantly reduces the complexity of searching for parallelism plans by leveraging two structural patterns found in large models. First, we identify parallel-preserving subgraphs, which ensure that the optimal global plan assigns the same parallel strategy to all operators within the subgraph. This approach allows us to avoid enumerating all possible combinations of parallel strategies for these operators. Second, we recognize repetitive subgraph patterns within the large computational graph, enabling us to profile a moderate number of representative subgraphs and accurately estimate the cost of parallelism plans with low overhead. With the significantly reduced search space, we can employ dynamic programming to search for the optimized parallelism plan. In our experiments, we demonstrate that CFP achieves significant speedups compared to the state-of-the-art framework for large models like GPT and LLAMA. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_00598 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | CFP: Efficient Optimization of Intra-Operator Parallelism Plans for Large Model Training Hu, Weifang Shi, Xuanhua Zhang, Yunkai Wu, Chang Peng, Xuan Zhai, Jiaqi Jin, Hai Qian, Xuehai Xue, Jingling Zhou, Yongluan Distributed, Parallel, and Cluster Computing Optimizing the parallel training of large models requires exploring intra-operator parallelism plans for a computation graph that typically contains tens of thousands of primitive operators. While the optimization of parallel data processing graphs has been extensively researched in database systems, the vast search space makes it challenging to apply traditional database query optimization methods and algorithms. This paper introduces CFP, an optimization system for intra-operator parallelism that significantly reduces the complexity of searching for parallelism plans by leveraging two structural patterns found in large models. First, we identify parallel-preserving subgraphs, which ensure that the optimal global plan assigns the same parallel strategy to all operators within the subgraph. This approach allows us to avoid enumerating all possible combinations of parallel strategies for these operators. Second, we recognize repetitive subgraph patterns within the large computational graph, enabling us to profile a moderate number of representative subgraphs and accurately estimate the cost of parallelism plans with low overhead. With the significantly reduced search space, we can employ dynamic programming to search for the optimized parallelism plan. In our experiments, we demonstrate that CFP achieves significant speedups compared to the state-of-the-art framework for large models like GPT and LLAMA. |
| title | CFP: Efficient Optimization of Intra-Operator Parallelism Plans for Large Model Training |
| topic | Distributed, Parallel, and Cluster Computing |
| url | https://arxiv.org/abs/2504.00598 |