https://github.com/geyang/ray-examples
https://github.com/geyang/ray-examples
Last synced: 3 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/geyang/ray-examples
- Owner: geyang
- Created: 2023-10-10T04:49:27.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2023-10-10T18:24:36.000Z (over 1 year ago)
- Last Synced: 2025-01-10T12:58:28.149Z (5 months ago)
- Language: Python
- Size: 199 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Ray Examples
I used this repo to learn how to write Async code with Pyro5, and then ray.
Pyro5's maintainer has made some bizzar decisions, and the project is practically dead. So after I wrote a version of our new project using Pyro4 (after migrating from Pyro5...), I decided to rewrite it using ray. The key resons for this rewrite are:
1. Pyro5 does not support pickle or cloudpickle serialization ([this guy is a douche](https://github.com/irmen/Pyro5/issues/16)).
2. Pyro4 does support those, but it does not support async function on the client side.
3. Ray supports both.
At last I needed to figure out how to setup a persistent Ray worker, so that I can fire up the LLM inference server and keep it running.
This is called a named actor in Ray.
See the code for details.
Ge