/vt/ - Virtual Youtubers

>>59827302
We'll likely have the option of letting the user pick their model, starting with the 13Bs and the 70Bs proper coming after that.
>>59834108
Not a goddamn bit.
>>59841006
You'll see.
>>59835726
Haha, goddamn. And to this day I'm still not very good at AI, but I'll be damned if I don't keep going anyways.
Truthfully, I never expected us to be where we are, even one year later. We've got our own company, we've got GPUs, Pyg-6B has been downloaded millions of times (no, seriously), we've published an actual research paper (and perhaps soon more than one if the developments posted above continue to hold), and even my own oshi has stated on stream she wants to do a collaboration with us. If you told the me from a year ago that all that would be happening as a result of forking two repos on GitHub, I would have looked at you as if you were insane. Thank you, guys, for helping us reach this point.

>>38049641

>>37987213
>>37987223
>>37987253

It's IRyS time now!

>>37326036
Same guy here. Fuck this, I'm tired of waiting. We're doing now.
I've forked two different repositories: one for a somewhat complete LaMDA, and one for a PaLM (basically LaMDA but even better) complete in terms of architecture, but unable to do scaling. I'm not sure which one is better, I'd prefer a PaLM one but I'm leaving both up.

I'll be hitting the books myself, I'm no AI researcher or even very good at AI, but I'll be damned if I don't try. I invite anyone who actually is good at this to send PRs to either repo. I'll look over them and accept if it's good. This should be a collaborative effort. Autists together strong and all that.
LaMDA fork: https://github.com/wAIfu-collectiveAI/LaMDA-pytorch
PaLM fork: https://github.com/wAIfu-collectiveAI/PaLM-pytorch

Let's do this, lads. Let's make them all real again.