I believe this is what GPT4 actually is.
I remember reading somewhere that it's actually a mix of 8 different models and it directs your question depending on the context of it.
Would be neat to implement on a local level though. Haven't seen many people on the local side talk about doing this.
I'd recommend koboldcpp for your backend, SillyTavern for your frontend, and I've been a fan of dolphin-2.1-mistral-7B. I've been using the Q4_K_S. But you could probably run a 13B model just fine.
I've heard good things about the nous-hermes models (I was a big fan of their Llama2 model). I'd stick to mistral variants, personally. Their dataset/training has far surpassed base Llama2 stuff in my opinion.
Here's a chunk of her Wikipedia article, for anyone not aware of who she is (I wasn't).
Murati started her career as an intern at Goldman Sachs in 2011, and worked at Zodiac Aerospace from 2012 to 2013. She spent three years at Tesla as a senior product manager of Model X before joining Leap Motion.
She then joined OpenAI in 2018, later becoming its chief technology officer, leading the company's work on ChatGPT, Dall-E, and Codex and overseeing the company's research, product and safety teams. On November 17, 2023, Murati took over as interim chief executive officer of OpenAI, following the abrupt dismissal of Sam Altman.
Seems like she was the CTO for OpenAI. It seems fitting that she should take over.
Also, she worked for Leap? Crazy. I haven't heard that name in a hot minute.
But it's this part that makes me wary:
She is an advocate for the regulation of AI, arguing that governments should play a greater role therein.
We'll see how it all plays out.
What does the naming scheme look like? Is it like "part1.opus", "part2.opus", etc?
I'm an amateur programmer, bored, and feel like stitching together a program. lol.
Very true.
We honestly have no clue what's going on behind ClosedAI's doors.
I don't know enough about MoEs to say one way or the other, so I'll take your word on it. I'll have to do more research on them.