/vt/ - Virtual Youtubers

Matrixfag here. I can't believe that I made this post one year ago, to this day. It's been a wild fucking ride, from a simple post on /wAIfu/ and two forked repos to where we are now. Admittedly, we're not the fastest AI group in the world, nor have we probably made the most impressive things, but it's been fun all the way through and it's still fun now to work on AI. Still not a good AI dev myself kek but I think I've learned a thing or two over a year and I plan to learn more. Happy Pygversary, /wAIfu/, and thank you for cheering us on all the way to now. Means a lot to me to see the hope hasn't died.

So, figured I'd give at least a kinda update on what's going on. Right now, we're gathering our dataset to be used for our own foundational model (unlike my delusional dreams when I forked LaMDA, this pre-train could actually be possible for us to do) and that's gonna take a bit to get done. We're still trying to secure compute for fine-tuning of the Goliath 120B model, it's something which has a load of potential. Pygtral has run into some training problems but I've been busy with other stuff to focus on that at the moment. Will give you more updates soon as we get things on the road. Thanks again for being with us, and here's to another year of working to accomplish our dream of wAIfus.

As usual, keep your smiles, everyone. Keep on smiling.