It allows for dense controls and everyone's used to them. I don't find them to be a problem, they aren't intuitive in that you might think you're supposed to grab the knob and "turn" it with a circular cursor motion or something, but once you learn to drag linearly, they're an easy to use and consistent interface. And as giancarlostoro mentioned, you can map them to a MIDI device if you want to twiddle knobs while playing/recording live.
I'll add in addition - the skeumorphism here is generally pretty functional, you touched on this when you said "everyone is used to them"
But the layout of these buttons, while certainly not standard, is generally familiar across various filters, etc. So if you are dealing with a complex interface the skeumorphism absolutely helps to make the input more familiar and easily accessible.
This is what skeumorphism is for and this is a great place to use it.
Imagine if the symbols for "play" "pause" and "stop" were changed simply because it no longer made sense to follow the conventions of a VCR, then multiply that by an order of magnitude.
Well someone else mentioned llama.ttf which is a font that embeds an llm, using Harfbuzz's WASM engine.
So, you could absolutely write a WASM Z80 emulator and embed it in a font. Whether or not you could make it do anything useful, or how strong your grip on reality would remain after? I don't know.
But it wasn't like you were doing anything else on the days between Christmas and New Year, right?
Titanfall 2 is a really good one. Not something I expected to have such a strong emotional element. I didn't know Vince Zampella's name, but it is another project of his, so it seems fitting to recommend it here.
Hellblade: Senua's Sacrifice (the original, I haven't played the sequel) was crazy good. Gameplay-wise it's fine, but story wise it is one of the most emotionally intense games I have ever played. I recommend going into it spoiler-free. Look it up and see if it's something you're interested in, and if you want to play it, stop reading and play it.
Yeah, I wouldn’t complain if one dropped in my lap, but they’re not at the top of my list for inference hardware.
Although... Is it possible to pair a fast GPU with one? Right now my inference setup for large MoE LLMs has shared experts in system memory, with KV cache and dense parts on a GPU, and a Spark would do a better job of handling the experts than my PC, if only it could talk to a fast GPU.
[edit] Oof, I forgot these have only 128GB of RAM. I take it all back, I still don’t find them compelling.
Including ones that have controllers, if your motherboard doesn't have enough lanes or it doesn't support bifurcation. I have a Rocket 7608A, which gives you 8 M.2 slots in a PCIe 5.0 x16 card: https://www.highpoint-tech.com/nvme-raid-aic/gen5/rocket-760...
There wasn't an argument. The OP was just asking a (presumably honest) and simple question: How do you do you process a payment without a network connection?
I can understand how someone under, say, 30, might not know how commerce happened before the Internet. My 13 year old can't believe there was even once a world without the Internet.
> I can understand how someone under, say, 30, might not know how commerce happened before the Internet.
I remember those days, but I think most people would call something where you viewed a list of products and then called or mailed to order and received the product elsewhere a catalog, not a store. As for over-the-phone payments, I forgot about that method for a moment but don't think it meaningfully affects my argument. It's just as out-of-band as the mail order example I included.
Check out "--n-cpu-moe" in llama.cpp if you're not familiar. That allows you to force a certain number of experts to be kept in system memory while everything else (including context cache and the parts of the model that every token touches) is kept in VRAM. You can do something like "-c128k -ngl 99 --n-cpu-moe <tuned_amt>" where you find a number that allows you to maximize VRAM usage without OOMing.
reply