The AI Horde is a project I started in order to provide access to Generative AI to everyone in the world, regardless of wealth and resources. The objective is to provide a truly open REST API that anyone is free integrate with for their own software and games and allows people to experiment without requiring online payment that is not always possible for everyone.
It is fully FOSS and relies on people volunteering their idle compute from their PCs. In exchange, you receive more priority for your own generations. We already have close to 100 workers, providing generations from stable diffusion to 70b LLM models!
Also the lemmy community is at !aihorde@lemmy.dbzer0.com
If you are interested in democratizing access to Generative AI, consider joining us!
Hey! Appreciate your post. The AI Horde has been one of my favorite projects to see evolve over the course of this year. Consider me subbed.
For myself (and others not as knowledgeable on the project), do you think you could briefly describe the main differences between how The AI Horde approaches crowd compute / inference compared to something like Petals? I know you mentioned here that the horde doesn’t do training. Is that the biggest difference to note?
Thanks again for your contribution to democratizing AI. Excited to see what The AI Horde can do with more supporters. I’ll be dedicating a few more nodes when I have a chance to spin them up.
Petals does distributed inference, which even with their enhanced system, is still pretty slow. The AI Horde sends the whole prompt to one individual worker and expects the reply back from that one worker. As such the speed is only dependent on the GPU that worker has. This limits us to how big models we can run, but then again I’ve seen people onboarding some crazy GPUs on the horde.
Appreciate the insight. I like that approach. I just learned you can become an Alchemist too. That’s a nice touch.
Alternatively you can become an Alchemist worker which is much more lightweight and can even run on CPU (i.e. without a GPU).
If you’re reading this, consider turning your PC into a Horde Worker! It’s a great way to contribute to AI (text & image-based) if you have the power to spare.
Out of pure curiosity: isn’t communicating via internet a bottleneck for distributed neural networks implementation? I mean the distribution of the work, not the API part
We’re not doing distributed inference. We’re doing distributed clustering The inference runs on individual PCs
So learning phase is distributed but then one prompt goes to one worker?
Yes, each prompt goes to a single worker. The AI Horde also doesn’t do training. We only do inference.
Does that mean that with a big (parameter-wise) model and not that powerful worker it can take a long time to respond?
Or the distributed clustering prevents from choking a worker?
Yes, it could. But we have a timeout. If a worker is unreasonably slow to respond, they get put into maintenance, so we expect people to only serve what they can run.
