Hetu is a high-performance distributed deep learning system targeting trillions of parameters DL model training, developed by DAIR Lab at Peking University. It takes account of both high availability in industry and innovation in academia.
This is the preview of Hetu 2.0, which is still under rapid development. Please raise an issue if you need any help.
We welcome everyone interested in machine learning or graph computing to contribute codes, create issues or pull requests. Please refer to Contribution Guide for more details.
-
Clone the repository.
-
Prepare the environment. We use Anaconda to manage packages. The following command create the conda environment to be used:
conda env create -f environment.yml. Please prepare Cuda toolkit, CuDNN, and gRPC in advance. -
We use CMake to compile Hetu. Please copy the example configuration for compilation by
cp cmake/config.example.cmake cmake/config.cmake. Users can modify the configuration file to enable/disable the compilation of each module. For advanced users (who not using the provided conda environment), the prerequisites for different modules in Hetu is listed in appendix.
# modify paths and configurations in cmake/config.cmake
# generate Makefile
mkdir build && cd build && cmake ..
# compile
# make hetu, version is specified in cmake/config.cmake
make -j 32- Prepare environment for running. Edit the hetu.exp file and set the environment path for python and the path for executable mpirun if necessary (for advanced users not using the provided conda environment). Then execute the command
source hetu.exp.
- Email: ccchengff@pku.edu.cn, xupeng.miao@pku.edu.cn
- Click here to join our Slack community.
- Hetu homepage: https://hetu-doc.readthedocs.io
- Committers & Contributors
- Contributing to Hetu
- Development plan
If you are enterprise users and find Hetu is useful in your work, please let us know, and we are glad to add your company logo here.
- Tencent Inc.
- Alibaba Cloud.
- Kuaishou Tech.
The entire codebase is under license
We have proposed numerous innovative optimization techniques around the Hetu system and published several papers, covering a variety of different model workloads and hardware environments.
- Xupeng Miao, Yujie Wang, Youhe Jiang, Chunan Shi, Xiaonan Nie, Hailin Zhang, Bin Cui. Galvatron: Efficient Transformer Training over Multiple GPUs Using Automatic Parallelism. VLDB 2023 [code]
- Youhe Jiang, Fangcheng Fu, Xupeng Miao, Xiaonan Nie and Bin Cui. OSDP: Optimal Sharded Data Parallel for Distributed Deep Learning. IJCAI 2023 [code]
- Yujie Wang, Youhe Jiang, Xupeng Miao, Fangcheng Fu, Xiaonan Nie, Bin Cui. Improving Automatic Parallel Training via Balanced Memory Workload Optimization. TKDE 2024 [code]
- Hao Ge, Fangcheng Fu, Haoyang Li, Xuanyu Wang, Sheng Lin, Yujie Wang, Xiaonan Nie, Hailin Zhang, Xupeng Miao, Bin Cui. Enabling Parallelism Hot Switching for Efficient Training of Large Language Models. SOSP 2024
- Xupeng Miao, Shenhan Zhu, Fangcheng Fu, Ziyu Guo, Zhi Yang, Yaofeng Tu, Zhihao Jia, Bin Cui. Reviving Efficient Attention for Long Context Language Modeling: A Survey. IJCAI 2024 [code]
- Pinxue Zhao, Hailin Zhang, Fangcheng Fu, Xiaonan Nie, Qibin Liu, Fang Yang, Yuanbo Peng, Dian Jiao, Shuaipeng Li, Jinbao Xue, Yangyu Tao, Bin Cui. Memo: Fine-grained Tensor Management For Ultra-long Context LLM Training. SIGMOD 2025
- Xiaonan Nie, Xupeng Miao, Zilong Wang, Jilong Xue and Lingxiao Ma, Zichao Yang, Gang Cao, Bin Cui. FlexMoE: Scaling Large-scale Sparse Pre-trained Model Training via Dynamic Device Placement. SIGMOD 2023
- Xiaonan Nie, Shijie Cao, Xupeng Miao, Lingxiao Ma, Jilong Xue, Youshan Miao, Zichao Yang, Zhi Yang, Bin Cui. EvoMoE: An Evolutional Mixture-of-Experts Training Framework via Dense-To-Sparse Gate. arXiv 2021 [code]
- Xiaonan Nie, Pinxue Zhao, Xupeng Miao, Tong Zhao, Bin Cui. HetuMoE: An Efficient Trillion-scale Mixture-of-Expert Distributed Training System. arXiv 2022 [code]
- Xiaonan Nie, Qibin Liu, Fangcheng Fu, Shenhan Zhu, Xupeng Miao, Xiaoyang Li, Yang Zhang, Shouda Liu, Bin Cui. LSH-MoE: Communication-efficient MoE Training via Locality-Sensitive Hashing. NeurIPS 2024
- Xupeng Miao, Hailin Zhang, Yining Shi, Xiaonan Nie, Zhi Yang, Yangyu Tao, Bin Cui. HET: Scaling out Huge Embedding Model Training via Cache-enabled Distributed Framework. VLDB 2022 (Best Scalable Data Science Paper) [code]
- Xupeng Miao, Yining Shi, Hailin Zhang, Xin Zhang, Xiaonan Nie, Zhi Yang, Bin Cui. HET-GMP: a Graph-based System Approach to Scaling Large Embedding Model Training. SIGMOD 2022 [code]
- Sicong Dong, Xupeng Miao, Pengkai Liu, Xin Wang, Bin Cui, Jianxin Li. HET-KG: Communication-Efficient Knowledge Graph Embedding Training via Hotness-Aware Cache. ICDE 2022
- Hailin Zhang, Penghao Zhao, Xupeng Miao, Yingxia Shao, Zirui Liu, Tong Yang, Bin Cui. Experimental Analysis of Large-scale Learnable Vector Storage Compression. VLDB 2024 [code]
- Hailin Zhang, Zirui Liu, Boxuan Chen, Yikai Zhao, Tong Zhao, Tong Yang, Bin Cui. CAFE: Towards Compact, Adaptive, and Fast Embedding for Large-scale Recommendation Models. SIGMOD 2024 [code]
- Zihao Yu, Haoyang Li, Fangcheng Fu, Xupeng Miao, Bin Cui. Accelerating Text-to-image Editing via Cache-enabled Sparse Diffusion Inference. AAAI 2024 [code]
- Xupeng Miao, Yujie Wang, Jia Shen, Yingxia Shao, Bin Cui. Graph Neural Network Training Acceleration over Multi-GPUs. Journal of Software (Chinese)
- Xupeng Miao, Xiaonan Nie, Yingxia Shao, Zhi Yang, Jiawei Jiang, Lingxiao Ma, Bin Cui. Heterogeneity-Aware Distributed Machine Learning Training via Partial Reduce. SIGMOD 2021
- Xupeng Miao, Yining Shi, Zhi Yang, Bin Cui, Zhihao Jia. SDPipe: A Semi-Decentralized Framework for Heterogeneity-aware Pipeline-parallel Training. VLDB 2023 [code]
- Xupeng Miao, Lingxiao Ma, Zhi Yang, Yingxia Shao, Bin Cui, Lele Yu, Jiawei Jiang. CuWide: Towards Efficient Flow-based Training for Sparse Wide Models on GPUs. TKDE 2021, ICDE 2021 [code]
- Xiaonan Nie, Xupeng Miao, Zhi Yang, Bin Cui. TSplit: Fine-grained GPU Memory Management for Efficient DNN Training via Tensor Splitting. ICDE 2022 [code]
- Xiaonan Nie, Yi Liu, Fangcheng Fu, Jinbao Xue, Dian Jiao, Xupeng Miao, Yangyu Tao, Bin Cui. Angel-PTM: A Scalable and Economical Large-scale Pre-training System in Tencent. VLDB 2023
If you use Hetu in a scientific publication, we would appreciate citations to the following papers:
@article{DBLP:journals/chinaf/MiaoXP22,
author = {Miao, Xupeng and Nie, Xiaonan and Zhang, Hailin and Zhao, Tong and Cui, Bin},
title = {Hetu: A highly efficient automatic parallel distributed deep learning system},
journal = {Sci. China Inf. Sci.},
url = {http://engine.scichina.com/doi/10.1007/s11432-022-3581-9},
doi = {10.1007/s11432-022-3581-9},
year = {2022},
}
@article{miao2021het,
title={HET: Scaling out Huge Embedding Model Training via Cache-enabled Distributed Framework},
author={Miao, Xupeng and Zhang, Hailin and Shi, Yining and Nie, Xiaonan and Yang, Zhi and Tao, Yangyu and Cui, Bin},
journal = {Proc. {VLDB} Endow.},
volume = {15},
number = {2},
pages = {312--320},
year = {2022},
publisher = {VLDB Endowment}
}
@article{ge2024enabling,
title={Enable Parallelism Hot Switching for Efficient Training of Large Language Models},
author={Ge, Hao and Fu, Fangcheng and Li, Haoyang and Wang, Xuanyu and Lin, Sheng and Wang, Yujie and Nie, Xiaonan and Zhang, Hailin and Miao, Xupeng and Cui, Bin},
journal = {Proceedings of the 30th {ACM} Symposium on Operating Systems Principles},
year = {2024},
publisher = {{ACM}}
}
We learned and borrowed insights from a few open source projects including TinyFlow, autodist, tf.distribute, FlexFlow and Angel.
The prerequisites for different modules in Hetu is listed as follows:
- OpenMP (*)
- CMake >= 3.24 (*)
- gRPC 1.6.3 (*)
- CUDA >= 11.8 (*)
- CUDNN >= 8.2 (*)
- MPI >= 4.1 (*)
- NCCL >= 2.19 (*)
- Pybind11 >= 2.6.2 (*)