https://github.com/zuston/horovod-yarn

Horovod on Yarn test case
https://github.com/zuston/horovod-yarn

Last synced: 25 days ago
JSON representation

Horovod on Yarn test case

• Host: GitHub
• URL: https://github.com/zuston/horovod-yarn
• Owner: zuston
• Created: 2021-04-04T02:49:18.000Z (almost 4 years ago)
• Default Branch: master
• Last Pushed: 2021-04-10T13:17:28.000Z (almost 4 years ago)
• Last Synced: 2024-10-17T05:55:56.621Z (3 months ago)
• Language: Python
• Size: 12.7 KB
• Stars: 2
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
Linked TonY(TF/PyTorch on Yarn) PR: https://github.com/linkedin/TonY/pull/524

## Horovod on Yarn

Aim to provide a Horovod local test program on Yarn, which can perform multi-process tests locally.

1. Driver start rendezvous server 

2. Inject some Horovod envs before starting training worker

__Driver__

```

python3 driver.py -w localhost:2

```

__Task__

```

python3 tensorflow2_minist.py --port=34824 --rank=0 --size=2 --local_rank=0 --local_size=2 --cross_rank=0 --cross_size=1  

```

```

python3 tensorflow2_minist.py --port=34824 --rank=1 --size=2 --local_rank=1 --local_size=2 --cross_rank=0 --cross_size=1

```