The recently discussed https://meshllm.cloud/ is the one I've been playing with but don't have the hardware to try with a model split between nodes, which apparently is supported and just not part of the public demo.
It's an interesting concept but the timing is probably too early. If more people had reliable low latency gigabit or ideally 10gbit throughput, then it might start to approach feasibility. There's other blockers too, but that springs to mind immediately.
It's cool to imagine a planet wide neural network interconnected with fiber - the nervous system of a planetary intelligence. But perhaps mushrooms do that already (Alpha Centauri ever relevant).
Petals is from 2022. Nowadays intelligence of smaller models, quantization techs, and optimizations to run models faster on consumer GPUs have improved a lot.
For distributed inference of smaller LLMs and diffusion models that fits in one consumer GPU rather than splits on multiple machines, there are already pretty good solutions such as AI Horde (formerly Stable Horde) [0]. Notably, it's the default provider that powers SillyTavern. It also has an interesting economy model of kudos.
> [...] volunteers can donate their hardware power to tackle a portion of a text-generating workload and team up others to complete larger tasks, similar to Folding@home and other distributed compute setups.
I have been thinking of something on these lines but with much smaller models. The entire model has to fit on a single computer. Host owner would choose the model they prefer, perhaps because they already use it. Then it is more about utilizing the GPU for LLM requests.
Peer to peer, consumers get to route their request to a host with compatible model. Consumers have to also contribute GPU but it does not have to be equal - I have not thought through the fairness part. Perhaps initially it starts with "create your friends group and have access to all the host nodes".
AI Horde has some measures to prevent Sybil attack that returns wrong results, but not enforce zero data retention. Prompts belong to the whole open source community. For example https://huggingface.co/datasets/la-ji/sd-prompt-in-the-wild
I do not know what/who Sybil is but yes, there is a lot of plumbing in order to make sure that host nodes are ZDR* compliant and much more. Zero visibility of source prompt.
Also, this is the reason I want to start with trust based groups only - invite people you already know. I have friends who have Macs with 32GB or 48GB unified memory but sending prompts to them is not a out of the box thing.
You can easily build a localized, access-controlled compute grid using Apple Silicon hardware. Haven't tested but as far as I can see, the key to making it performant would be keeping full models resident on individual nodes rather than attempting to distribute the tensor processing across the network.
Any updates here to make this work on more modern llms?
The recently discussed https://meshllm.cloud/ is the one I've been playing with but don't have the hardware to try with a model split between nodes, which apparently is supported and just not part of the public demo.
Can't even scroll the website on mobile, too much going on
It's an interesting concept but the timing is probably too early. If more people had reliable low latency gigabit or ideally 10gbit throughput, then it might start to approach feasibility. There's other blockers too, but that springs to mind immediately.
It's cool to imagine a planet wide neural network interconnected with fiber - the nervous system of a planetary intelligence. But perhaps mushrooms do that already (Alpha Centauri ever relevant).
Relevant: Why Switzerland has 25 Gbit internet and America doesn't https://news.ycombinator.com/item?id=47652400
Too early too in that later (hopefully) each node can serve a full model so it is then a simpler share tit for tat of complete models.
Too early? the project has been going for years now...
Petals is from 2022. Nowadays intelligence of smaller models, quantization techs, and optimizations to run models faster on consumer GPUs have improved a lot.
For distributed inference of smaller LLMs and diffusion models that fits in one consumer GPU rather than splits on multiple machines, there are already pretty good solutions such as AI Horde (formerly Stable Horde) [0]. Notably, it's the default provider that powers SillyTavern. It also has an interesting economy model of kudos.
[0] https://stablehorde.net/
Seems a bit like GNUS [0].
---
[0]: https://www.gnus.ai/
Hm, looks like this is an old project (2022) associated with HuggingFace (https://huggingface.co/bigscience).
Doesn't look like it's very active nowadays: https://github.com/bigscience-workshop/petals
Article with more details: https://techcrunch.com/2022/12/20/petals-is-creating-a-free-...
> [...] volunteers can donate their hardware power to tackle a portion of a text-generating workload and team up others to complete larger tasks, similar to Folding@home and other distributed compute setups.
I have been thinking of something on these lines but with much smaller models. The entire model has to fit on a single computer. Host owner would choose the model they prefer, perhaps because they already use it. Then it is more about utilizing the GPU for LLM requests.
Peer to peer, consumers get to route their request to a host with compatible model. Consumers have to also contribute GPU but it does not have to be equal - I have not thought through the fairness part. Perhaps initially it starts with "create your friends group and have access to all the host nodes".
Can't wait to run a Sybil farm of subsidized proxy nodes and log all the juicy prompts.
AI Horde has some measures to prevent Sybil attack that returns wrong results, but not enforce zero data retention. Prompts belong to the whole open source community. For example https://huggingface.co/datasets/la-ji/sd-prompt-in-the-wild
I do not know what/who Sybil is but yes, there is a lot of plumbing in order to make sure that host nodes are ZDR* compliant and much more. Zero visibility of source prompt.
Also, this is the reason I want to start with trust based groups only - invite people you already know. I have friends who have Macs with 32GB or 48GB unified memory but sending prompts to them is not a out of the box thing.
* typo
Imagine if there was some kind of way to cryptographically 'prove' your GPU is doing some kind of 'work' here and contributing to the network.
You could even hand out some kind of digital 'currency' to the people proportional to the amount of work their doing!
As we've seen, that's a solution looking for a problem.
But what if the work is useful beyond just crunching numbers for the sake of proving it?
Not gonna work, will be starved by inter node bandwidth.
You can easily build a localized, access-controlled compute grid using Apple Silicon hardware. Haven't tested but as far as I can see, the key to making it performant would be keeping full models resident on individual nodes rather than attempting to distribute the tensor processing across the network.