someone else closely involved with SNet or the authoring of the paper on Hyperon. It is described that there are plans for multiple demonstration models during the alpha launch in March, for instance, a Hyperon-based framework for image generation. It's exciting to see that we are finally witnessing some real applications. Such models surely require substantial processing power and many GPUs for training. Are there external investors supporting this, or is it entirely funded by SNet internally? Given that Hyperon is launching in a decentralized manner, I imagine the financial aspects of training these models might be complex?"
Anyone from the team able to comment on this?
I'm not from the team, but NTX pops up in my mind when reading that.
Understood, but the resources that can be quite heavy still need to be paid. Nunet or centralized
Yes, decentralised computing paid in NTX. We'd need the AGIX marketplace to create this revenue of course.
From the devs: A short answer is that we are not training new models, but wrap open-sourced models. Well, we do train small components of the models for image generation, which don't require a lot of GPUs. Som other models are also trained by Sergey Shalyapin, in particular for NLP. And there are plans to get more GPUs through Zarqa, but basically the question is based on a confusion.
Understand! The question was indeed based on a confusion. So the models will also indeed been launched on snet with the alpha? As i read the paper and understand the end goal to launch everything on Snet in 18-24 months. But it was not really clear to me if its also the goal the launche it on snet from the alpha release in March?
Обсуждают сегодня