Added blog post for announcement - draft #76
Conversation
There was a problem hiding this comment.
@filopedraz lemme know if I should go ahead and commit these minor changes that I added, also had few doubt questions above.
Also what kind of improvements are you looking to be added here? like the section Conclusion is not fully filled, so those? and maybe fill up other less filled sections?
|
|
||
| > Horizontal Scaling is all you need. | ||
|
|
||
| Prem Labs is introducing **Prem Network**, a decentralized AI Platform enabling consumer devices to run Large Language Models without compromising on quality, aggregating the computational power of devices around the globe. Prem Network is a breakthrough in LLM inference deployment as it enables consumer devices to power GPU-intensive models on the edge. Similar to what BitTorrent did for P2P file sharing, Prem Swarm provides AI inference and training for consumer devices. |
There was a problem hiding this comment.
@filopedraz what's the difference between Prem Network and Prem Swarm? I am seeing swarm mentioned in the end of this para whereas it's all about network before
There was a problem hiding this comment.
Prem Swarm is the old name. We should replace it everywhere with Prem Network.
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
| - Context dimension | ||
| - Generation quality and hallucination | ||
|
|
||
| Some very good papers have come out lately to face these issues (e.g., X, Y), but no major breakthrough has yet emerged. It would seem that perhaps the Transformer model architecture optimization itself is not good enough for the next generation of LLMs with consumer hardware constraints. Only research will be able to answer these open questions. |
There was a problem hiding this comment.
Some very good papers have come out lately to face these issues (e.g., X, Y)
Do we have a list already for this? or should I lookup/find few?
There was a problem hiding this comment.
No list. Two options: (1) we look them up, (2) we remove this strong sentence. Wdyt?
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
blog/2023-11-06-towards-censorship-resistant-architecture/index.md
Outdated
Show resolved
Hide resolved
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
…x.md Co-authored-by: Biswaroop <biswaroop08@gmail.com>
|
|
||
| > Horizontal Scaling is all you need. | ||
|
|
||
| Prem Labs is introducing **Prem Network**, a decentralized AI Platform enabling consumer devices to run Large Language Models without compromising on quality, aggregating the computational power of devices around the globe. Prem Network is a breakthrough in LLM inference deployment as it enables consumer devices to power GPU-intensive models on the edge. Similar to what BitTorrent did for P2P file sharing, Prem Swarm provides AI inference and training for consumer devices. |
There was a problem hiding this comment.
@filopedraz is it only for LLM or could it be more general (text-to-image, text-to-video, ...)?
There was a problem hiding this comment.
For now only LLM.
There was a problem hiding this comment.
@filopedraz should we add references to State of Open Source AI Book - 2023 Edition?
There was a problem hiding this comment.
Good point. Yes. We should, but where would you put it?
There was a problem hiding this comment.
We can either use it to introduce Prem and its contribution to open source trough the book or reference chapters when we talk about fine tuning, open-source and so on. Or even both if it's not too much
No description provided.