Obviously there’s not a lot of love for OpenAI and other corporate API generative AI here, but how does the community feel about self hosted models? Especially stuff like the Linux Foundation’s Open Model Initiative?
I feel like a lot of people just don’t know there are Apache/CC-BY-NC licensed “AI” they can run on sane desktops, right now, that are incredible. I’m thinking of the most recent Command-R, specifically. I can run it on one GPU, and it blows expensive API models away, and it’s mine to use.
And there are efforts to kill the power cost of inference and training with stuff like matrix-multiplication free models, open source and legally licensed datasets, cheap training… and OpenAI and such want to shut down all of this because it breaks their monopoly, where they can just outspend everyone scaling , stealiing data and destroying the planet. And it’s actually a threat to them.
Again, I feel like corporate social media vs fediverse is a good anology, where one is kinda destroying the planet and the other, while still niche, problematic and a WIP, kills a lot of the downsides.
What RAM capacity?
Honestly, if LLMs are your focus, you should just upgrade to a used M2 Max (or Ultra) when the M4 comes out, lol. Basically the only thing that matters is RAM capacity and bandwidth, and the M2 is just going to be faster and better than a similarly priced M4.
Or better yet, upgrade to and AMD Strix Halo. This will buy you into linux and the cuda ecosystem (through AMD rocm), which is going to open a lot of doors and save headaches (while admittedly creating other headaches).
Honestly I’ve mostly been playing around with image generators and learning how to write a good prompt. But LLMs are where I see real value and would love to learn more about self-hosting one, and custom training it.
I find it interesting that the M2 is better for LLM than the M4 will be… what’s the reason for this?
RAM capacity and bandwidth.
That basically the only two things that matter for local LLM performance, as it has to read the entire model from memory for every token (aka half word). And for the same money, a “higher end” M2 (like an M2 Max or Ultra) will just have more of it than the equivalent cost M3 or (probably) M4.