Yes, you can probably get the whole thing up and running in about an hour the first time. If you pause and restart, it takes about 15 minutes to load the models from disk into GPU memory, so budget for cold startup time.
How does it take fifteen minutes to read <100 GB into GPU memory? Shouldn't that be limited by SSD speed with everything slower than a minute being a terrible ssd?
A lot of cloud platforms have terrible slow network storage. They also might need to compile the GPU kernels fresh as they might not have a persistent CUDA cache.
this is not right. vllm can easily load safetensors model at >5GB/s if not faster when setup right. did you make the compile cache persist? if you use docker, you should bind mount the kernel compile cache, so they don't need to be recompiled each time vllm restart.
It tells a specific counterfactual story, but it's one that has a reasonable background for it.
If you were a new computer user in 1980, your Appie II+ or PET 8032 booted straight into BASIC, and it was a good-enough tool that could do productive development for the entire life of the machine there.
But if you were a new computer user in 1990, you didn't have that. They gave you GW-BASIC, and if you're lucky a separate manual rather than a vague chapter in the main DOS manual. This might be enough to give you some initial experience with programming, but you have a bunch of hard brick walls.
Your machine might have 640k-4Mb of memory, but GW-BASIC can only meaningfully work with 64k of it.
You can't access the mouse or any post-EGA graphics modes without a LOT of assembly shimming. If you had a sound card, it was outside the built-in tooling.
Remember, this is a few years before the Internet being widely available, and stuff like the GNU project being widely accessible. This is probably the best development tool you have available
If you wanted to go further in programming, it meant buying a boxed compiler that probably cost well in excess of USD100, and was for a completely different language than the one you had cut your teeth on.
But if you wanted "The GW-BASIC I already understand, but more", you want this.
It was not the future of the language. It's starting, as the blurb says, from where GW-BASIC and the old Microsoft BASIC were in the early 1980s. We know what the actual future of BASIC was after those, because that future happened. It wasn't this at all.
Was the name change related to publicity from that? My memory is that it was part of an attempt to be the new Second Life. But on the other hand I tend to be naive.
Yes, they did it to break the association with their brand (“reputation laundering”) and then they got into 3D to throw some chaff. Seems to have worked for them.
If they changed their name to “Flobular” it wouldn’t change a thing.
They’re Facebook in many people’s minds.
“Meta” doesn’t convey the same meaning that “Facebook” does. And it’s also squatting on an important term in English. Mets is used all over the language, like metaphysical.
Right on, I agree with you on all points. I just found it easy enough to change that last letter to a 't'.
Because in addition to simply opposing corpos' name resets, I'll go even further by modifying their names to better label their behavior. Corpos spend gobs of cash on naming and branding - it must mean something. So why not embrace that dynamic from an individualist perspective?
> Nvidia is literally paying for datacenter construction, which means paying other companies to buy their chips.
Does Nvidia pay more than it receives from datacenter construction?
> A great website to bookmark and keep an eye on is: isaiprofitable. At the bottom, you’ll see that Nvidia (selling pans, shovels, and blue jeans) is the only company making money.
Have you considered that:
1) There are many other companies making a lot of money? Examples: NVIDIA, Micron, TSMC, Broadcom, AMD, SK Hynix, ASML...
2) The hyperscalers (Google, Microsoft, Amazon) are profitable?
Honestly the word think is so poorly defined it cannot be used with any scientific rigor and is rather useless without a dissertation being posed with it on what you actually mean by that. Intelligence really needs one too.
> I'm not sure about coders but for artists I don't think a massive number were making anime waifus, but instead were making graphic designs for advertisements, publications & websites.
There was an old joke among designers that their biggest competitor was the boss' nephew playing with photoshop.
Now they need to compete with the boss' nephew playing with AI.
Technically, AI can introduce rounds of fact-checking, proofreading, etc.
But even without that, AI-generated text is already better than what most humans can write -- particularly when they need something fast, cheap and good.
(It might not be "excellent" -- but humans wouldn't be able to make it fast and cheap...)
Of course if the text is made with care and re-read by a human it will be okay even if ai generated.
But I frequently encounter two issues:
1. Careless text generation, that was never read and so the human never prompted the LLm for proofreading and so on, this results in a text/readme/documentation that is using way too much words to describe simple or mildly complex things
2. LinkedIn post/corporate-trying-to-seem-friendly type of text that is always writing like it has to convince you of something (frequent user of the "it's not just X - It's y" trope)
I find that for most AI generated texts, it defaults to my 2nd point, and the human often doesn't read it or is happy with it, whereas I would have been much happier reading the shorter version they would have written instead.
Is it viable to start/stop it multiple times per day?