andersa

10 ms·

121.
by andersa 2y ago
I see. This sounds like a problem for people using public wifi...
122.
by andersa 2y ago
But why would such desktops be exposed to the public internet directly?
123.
by andersa 2y ago
So just to make sure I understand correctly, this is a nothingburger, right? No important server has a printer attached. Any basic firewall would block this traffic.
124.
by andersa 2y ago
Not before the bubble pops.
125.
by andersa 2y ago
Hmm. Seems like a bad deal. This is a monthly reservation for a single 6000 Ada for $940. You can get the same on RunPod for $670. And to actually train stuff you'd likely want nodes with more of them, like 8, or just different GPUs al
126.
by andersa 2y ago
Just saw this announced on Twitter. https://x.com/__tinygrad__/status/1838508142415737133
127.
Tinybox Pro with 8x 4090(tinygrad.org)
1 points by andersa 2y ago|1 comments
128.
by andersa 2y ago
No. That's why a lot of people are scared of it.
129.
by andersa 2y ago
This would be unbelievably inefficient, game engines will be running hundreds of millions of functions per second. And if the code runs 10x slower with the trace active, then it's no longer sensible. We use sampling for the cases where
130.
by andersa 2y ago
This is really suboptimal. They should buy 4 more 3090s so they can run it massively faster with tensor parallelism.
131.
by andersa 2y ago
If you really want to know an exact number for a specific use case, you can rent an 8xH100 node on RunPod and benchmark it. You should expect somewhere around 30t/s for a single response, if running the FP8 rowwise quant that would typ
132.
by andersa 2y ago
You missed the part where this is slow as hell.
133.
by andersa 2y ago
It is actually much more expensive than 20k. I can find them between 25-30k only.
134.
by andersa 2y ago
They don't prescribe anything, it's just there. Like a box of sweets at the supermarket. Or a bottle of beer. Or whatever else people like.
135.
by andersa 2y ago
People are often scared of things they don't understand.
136.
by andersa 2y ago
Unlike with a real person, where you only find this out much later, here you know the connection is shallow and meaningless from the start. That seems like an upgrade! Now you can have fun with no consequences.
137.
by andersa 2y ago
This doesn't even read like a real conversation... chatbots can do a lot better than this
138.
by andersa 2y ago
What makes you think they can't do that?
139.
by andersa 2y ago
AI mostly does whatever you ask it to. You can easily make a chatbot that will question your choices or act like a hater on twitter.
140.
by andersa 2y ago
> I can make the case much more convincingly that food has basically no intrinsic added value to your life and if most people could press a button to make themselves never eat again, they would. What nonsense is this? Do you only eat bea
141.
by andersa 2y ago
The goal is not to develop social skills. The goal is to not need them, without being lonely.
142.
by andersa 2y ago
They never intended to be open or share any of their impactful research. It was a trick the entire time to attract talent. The emails they shared as part of the Elon Musk debacle prove this: https://openai.com/index/ope
143.
by andersa 2y ago
Do people really use text search for this rather than an IDE that parses all of the code and knows exactly where each declaration is, able to instantly jump to them from a key press on any usage...? Wild.
144.
by andersa 2y ago
So if it doesn't use 3.2kW continuously, but varies significantly based on what it's doing (perhaps even idle sometimes) then it's fine?
145.
by andersa 2y ago
Usually you want to split each layer to run with tensor parallelism, which works optimally if you can assign each kv head to a specific GPU. All currently popular models have a power of 2 number of kv heads.
146.
by andersa 2y ago
I don't think they're intended for rack usage like that. More like for people to put under their desks... there would be no reason to build the giant case with fancy silent-ish cooling if you're going to put them next to your
147.
by andersa 2y ago
Most models actually can't be split nicely by 6. There's a reason nvidia builds nodes with 4 and 8 GPUs.
148.
by andersa 2y ago
It's for personal use, though it would not run any games, it would be for running offline inference and other experiments. Probably not a smart purchase, but a fun one... That is good to know multiple phases can work. Perhaps there wou
149.
by andersa 2y ago
Motherboard and CPU, memory, NVMe drives, PSUs, SlimSAS cables and breakouts, a custom machined case, assembly, support. You're free to try building one yourself for cheaper. If you consider your time for researching/assembling&#x
150.
by andersa 2y ago
Why is it a fire hazard if the extension cords are properly rated for this load?

More ›