search

redotvideo / haven

LLM fine-tuning and eval
https://haven.run
Apache License 2.0
341 stars 11 forks source link

Is this maintained? #84

Open osilverstein opened 1 year ago

osilverstein commented 1 year ago

it should be

getting loss of 0 after following instructions. slowly dropped to... 0? 128, 64

sqpollen commented 1 year ago

No replies in any question on the Discord

justusmattern27 commented 1 year ago

Hey guys! Sorry for being unresponsive on Discord. We had a lot of stuff to do over the last weeks. Regarding your question:

We have decided to move on from maintaining this GitHub repository - particularly, our decision was driven by your feedback on the following points:

  1. The setup process of this project does not provide the desired simplicity for an LLM deployment tool. Ideally, local deployment should be as simple as using serverless platforms like Replicate, where users can log in to the website, choose a model they want to run, and deploy it with the click of a button.

  2. This project does resource management and orchestration through our own custom manager service - this is something that Kubernetes does out of the box, and much better than we can do it. To replace our manager with Kubernetes, we would essentially have to rewrite the entire codebase.

To address the problems described, we have built a web-based offering of Haven, which strongly simplifies the setup process and is built on top of Kubernetes. You can find it on our website. We'd be really happy to hear what you think and are super thankful for the valuable feedback and support we received for this repository 💜. Please contact us on hello@haven.run if you have any questions!

uniAIDevs commented 8 months ago

I'm working on a duplicate version of this app that will be able to be self-deployed. It is extremely complex but I have good work done so far. I will share the repo to this thread once I am done. I also have AI agents working 24/7 to speed up this dev process. I hope to be finished within the week!