https://github.com/d4l3k/axe

A simple graph partitioning algorithm written in Go. Designed for use for partitioning neural networks across multiple devices which has an added cost when crossing device boundaries.
https://github.com/d4l3k/axe

graph-partitioning machine-learning model-parallelism

Last synced: about 1 year ago
JSON representation

A simple graph partitioning algorithm written in Go. Designed for use for partitioning neural networks across multiple devices which has an added cost when crossing device boundaries.

• Host: GitHub
• URL: https://github.com/d4l3k/axe
• Owner: d4l3k
• License: mit
• Created: 2020-06-13T22:32:03.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2020-06-17T17:43:03.000Z (about 6 years ago)
• Last Synced: 2025-02-08T09:30:33.631Z (over 1 year ago)
• Topics: graph-partitioning, machine-learning, model-parallelism
• Language: Go
• Size: 9.77 KB
• Stars: 1
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# axe

A simple graph partitioning algorithm written in Go.

Designed for use for partitioning neural networks across multiple devices which has an added cost when crossing device boundaries.

Current algorithm uses a form of simulated anealing to find an approximate global minima and then once the temperature is 0 does greedy optimizination to find the local minima.

## License

Copyright (c) 2020 Tristan Rice

Licensed under the MIT license.