It is! 143GB last I checked. I’m on 128GB RAM + 3090, 1 NUMA node, so I think it’s juuust barely too tight. But it should be perfect with a few of the “sparsest” MoEs quantized.
If KTransformers supports something like that, I may have to finally check it out, since v4 won’t need many esoteric features.
Not to speak of the new attention scheme and the (IIRC) MLP changes.
I’m very much looking forward to ik_llama.cpp implementing it. I don’t think I can quite fit Flash on my rig (hence no Ktransformers for me) but a little quantization of the sparse layers, and it’d be perfect.
This is commonly cited, but not strictly true.
Prompt processing is completely compute limited. And at high batch sizes, where the weights are read once for many tokens generated in parallel, token generation is also quite compute limited. Obviously you want enough bandwidth to match the compute, but its very compute heavy.
You can see this for yourself. Try ~10 prompts in parallel on a CPU in llama.cpp, and it will slow to a crawl, while a GPU with a narrow bus won’t slow down much.
Training is a bit more complicated, but that’s not doable on CPUs anyway.
Now, local inference (aka a batch size of 1), past prompt processing, is heavily bandwidth limited. This is why hybrid inference works alright on CPUs. But this doesn’t really apply to servers, which process many users in parallel with each “pass”.
I’ve heard hints they sometimes give GPU farms busywork, to meet utilization targets apparently stipulated in sales contracts.
No. Not even close. Non-US models are trained (and run) on peanuts compared to big US models, because they don’t have mega GPU farms and have no other option. Deepseek in particular went all-in on software architecture efficiency.
…Ironically, the Nvidia GPU embargo was the best thing that ever happened to the Chinese devs. It made them thrifty.
Many tried to warn US regulators of this, but they had AI Bros whispering in their ears. The US tech system is just too screwed up, I guess.
…I mean, yeah? It’s obviously developed for the Chinese market.
But that’s theoretical, for now. No CPU backend I can find supports DSV4, and DeepSeek hasn’t contributed anything yet.
Just not power/cost efficiently on CPU only, is what I meant. CPUs don’t have the compute for batching (running generation requests in parallel). You need an accelerator, like Huawei’s, to be economical.
It’s fine for local inference, of course.
I just meant for mass inference serving.
Yeah, I haven’t seen much in the way of bitnet training savings yet, like regular old QAT. It does appear that Deepseek is finetuning their MoEs in a 4-bit format now, though.
What’s left unsaid is the software architecture is extremely interesting, and efficient.
Ironically, the Nvidia embargo was the best thing to ever happen to the Chinese labs (which Nvidia tried to tell the US govt). It forced them to get thrifty, unlike US labs which (allegedly) fill some GPU farms with busywork for the appearance of high utilization.
Not at scale. Even on the new architecture, one really needs some kind of accelerator to make it economical for servers.
Bitnet-like models might change the calculus, but no major trainer had tried that yet.
Also, I was thinking of various wallet drainers, not just simple transactions or classic scams: https://www.kaspersky.com/blog/what-is-a-crypto-wallet-drainer/50490/
Etherium isn’t just a currency, but a contract system. You basically have to be a digital lawyer+developer to understand it, and not accidentally click on the wrong thing to drain your account, and that is a LOT to ask of the average person.
We’re on Lemmy, which must put us in the first percentile of tech nerds, yet I don’t have the bandwidth to learn enough of that to feel secure with my savings in an Etherium wallet. How could the average person?
Anarchist types are concerned about government backed crypto coins since you lose the fungibility/anonymity of physical dollars but don’t get any of the freedom and separation from centralization that crypto supposedly represents.
Plus all the potential for oligarch corruption, like current crypto has. Yeah, it’s like the worst of everything, by design.
But Etherium isn’t protected from that, either.
I like the idea of decentralized digital currency insulated from governments, I just feel the Etherium-style blockchain approach is just impractical, and not that, on so many levels.
Were they robbed?
I bet they were.
Say what you will about cash, but some hacker isn’t taking paper bills from across the planet via some technical exploit way over my head. With Etherium, the only thing protecting your money from the entire internet is you, and your understanding of complicated intricacies… And when lost, no one is coming to help you.
They might get my credit card, yeah. But that’s either my own dumb fault, or a very rich bank’s problem.
…It’s great for scammers, though. Crypto’s like a wet dream for them. And I find it remarkable the crypto community sees that as a feature, not a bug, and somehow thinks the whole world must see it that way.
I mean this politely but… read the article?
I mean, I use every alternative I can. Vapoursynth scripts, libraw-based projects, random GitHub repos, DaVinci…
But there are some features I just can’t get great support for outside of definitely-not-high-seas Lightroom Classic:
Good lens profiles for weird lenses.
Proper HDR PQ/HLG editing and AVIF/JXL export support.
RAW support for newer cameras, like my little R50V
I have yet to try DaVinci’s photo editing mode though. That’s very interesting.
“If I wish for three more wishes, you will grant them with no catches.”
Keep doing this, over and over again, trying different strategies. It’s a way to test and validate genie loopholes (as you will be unable to state that if it’s wrong).
Though if the genie was smart, being “wrong” would be qualified to your own internal knowledge, I suppose. E.g. you can’t knowingly lie.
On a technical level, that makes zero sense.
AI “agents” are basically just fancy prompts with a tool calling harness. They are infinitely replicable, at zero cost, with no intrinsic value; the cost comes from the generic CPU host, and the API calls to GPU servers, databases, or whatever else that are all centralized anyway.
Wanna hear a dirty secret?
“AI” cost is going to zero.
Model capabilities aren’t scaling, but inference efficiency is exploding, thanks to more resource-constrained labs and breakthroughs in papers. The endgame of the current bubble is mediocre but useful tools anyone can host themselves, dirt cheap. Maybe a bit more reliable and refined than what we have now, but about as “intelligent.”
And guess what?
Microsoft can’t profit off that. None of the Tech Bros can.
Point being, this exec is either delusional, or jawboning, so the world doesn’t realize that “AI” is a dumb utility/aid, and they can’t make any profit off it.
Regardless of whatever fraction most of the revenue comes from, they still draw absolutely massive amounts of players.
I think the problem is at the other end: the ads.
And platforms.
Some AI ad of Tom Hanks peddling a supplement, or a sexy ad of AI Taylor Swift, shouldn’t be distributed en masse in the first place, just because an algorithm or ad engine picked it up as engagement bait. It’s insane! There is nothing normal about it, and its about time we stop pretending the screwed up platforms profiting off this stuff are “free speech” and acceptable.
…Because scammers are always gonna scam. But they can only do this because the platforms are pourinf fuel on the fire.