ChocoSGD
所属分类：论文
开发工具：Python
文件大小：342KB
下载次数：0
上传日期：2020-09-10 21:44:26
上 传 者：
sh-1993
说明： 去中心化SGD与通信压缩共识：<https://arxiv.org abs 1907.09356>
文件列表:
LICENSE (11357, 2020-09-11)
convex_code (0, 2020-09-11)
convex_code\base_logistic.py (2632, 2020-09-11)
convex_code\baselines.py (994, 2020-09-11)
convex_code\constants.py (43, 2020-09-11)
convex_code\dump (0, 2020-09-11)
convex_code\dump\optimum-epsilon (0, 2020-09-11)
convex_code\dump\optimum-epsilon\baselines.pickle (135, 2020-09-11)
convex_code\experiment.py (981, 2020-09-11)
convex_code\experiment_epsilon_final.py (15320, 2020-09-11)
convex_code\logistic.py (8603, 2020-09-11)
convex_code\parameters.py (4683, 2020-09-11)
convex_code\pickle_datasets.py (639, 2020-09-11)
convex_code\plots-epsilon.ipynb (303279, 2020-09-11)
convex_code\utils.py (550, 2020-09-11)
dl_code (0, 2020-09-11)
dl_code\auto_extract.py (1004, 2020-09-11)
dl_code\environments (0, 2020-09-11)
dl_code\environments\docker (0, 2020-09-11)
dl_code\environments\docker\base (0, 2020-09-11)
dl_code\environments\docker\base\.screenrc (2107, 2020-09-11)
dl_code\environments\docker\base\.tmux.conf (1628, 2020-09-11)
dl_code\environments\docker\base\Dockerfile (3372, 2020-09-11)
dl_code\environments\docker\base\entrypoint.sh (33, 2020-09-11)
dl_code\environments\docker\base\fix-permissions (963, 2020-09-11)
dl_code\environments\docker\docker-compose.yml (309, 2020-09-11)
dl_code\environments\docker\pytorch-mpi (0, 2020-09-11)
dl_code\environments\docker\pytorch-mpi\Dockerfile (2915, 2020-09-11)
dl_code\exps (0, 2020-09-11)
dl_code\exps\finetune-lstm (0, 2020-09-11)
dl_code\exps\finetune-lstm\example.sh (1110, 2020-09-11)
dl_code\exps\finetune-resnet20-cifar10 (0, 2020-09-11)
dl_code\exps\finetune-resnet20-cifar10\example.sh (6175, 2020-09-11)
dl_code\exps\finetune-resnet50-imagenet (0, 2020-09-11)
dl_code\exps\finetune-resnet50-imagenet\decentralized_sgd.sh (1361, 2020-09-11)
... ...
# Choco-SGD
This repository provides code for **communication-efficient decentralized ML training** (both deep learning, compatible with [PyTorch](https://pytorch.org/), and traditional convex machine learning models.
We provide code for the main experiments in the papers
- [Decentralized Stochastic Optimization and Gossip Algorithms with Compressed Communication](https://arxiv.org/abs/1902.00340) and
- [Decentralized Deep Learning with Arbitrary Communication Compression](https://arxiv.org/abs/1907.09356).
Please refer to the folders `convex_code` and `dl_code` for more details.
# References
If you use the code, please cite the following papers:
```
@inproceedings{koloskova2019choco,
title = {Decentralized Stochastic Optimization and Gossip Algorithms with Compressed Communication},
author = {Anastasia Koloskova and Sebastian U. Stich and Martin Jaggi},
booktitle = {ICML 2019 - Proceedings of the 36th International Conference on Machine Learning},
url = {http://proceedings.mlr.press/v97/koloskova19a.html},
publisher = {PMLR},
volume = {97},
pages = {3479--3487},
year = {2019}
}
```
and
```
@inproceedings{koloskova2020decentralized,
title={Decentralized Deep Learning with Arbitrary Communication Compression},
author={Anastasia Koloskova* and Tao Lin* and Sebastian U Stich and Martin Jaggi},
booktitle={ICLR 2020 - International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=SkgGCkrKvH}
}
```
近期下载者：
相关文件：
收藏者：