terafo

12 ms·

121.
by terafo 4y ago
Actual code of tinygrad is less than 5k lines. There is also 1600 lines of tests and around 2k lines of example models. And I didn't count unfinished support for geohot's own unfinished neural network accelerator(verilog for that
122.
by terafo 4y ago
I would argue that there are performance *benefits* for a developer in running python code, due to how programs are run in python(Jupyter notebooks) you basically can change program on the fly, and not recompile and restart it, as you would
123.
by terafo 4y ago
It wouldn't be huge, I once recorded a week of me using my pc(so around 80 hours in total), and it was sub-100 gigs. It was 1080p with decent quality, don't remember FPS though.
124.
by terafo 4y ago
It isn't 19.5 hour stream, I went and randomly clicked on couple of timestamps and stumbled upon[1]. So it's two almost-10-hour-long streams put together because they are thematically similar. [1] https://youtu.be/
125.
by terafo 4y ago
I think it's 300 million[1], not billion. [1] https://www.clintonfoundation.org/wp-content/uploads/2021/11...
126.
by terafo 4y ago
Regular people aren't buying ASML equipment either. Aren't customers of Maersk. Wouldn't buy A330. They benefit from their products every day, though. And yeah, regular people are increasingly using Starlink, especially in my
127.
by terafo 4y ago
Are there any ferries that can hit 200 km/h? Because that's what it would take to get reasonably far from the city.
128.
by terafo 4y ago
You are plainly wrong. Out of 50 Falcon 9 launches this year only 3 had DoD payloads.
129.
by terafo 4y ago
A lot of things happened. Apollo 1 disaster, end of space race, space shuttle not hitting enough launch volume and having unclear requirements, Challenger and Columbia incidents which skyrocketed shuttle between-flight maintenance, continue
130.
by terafo 4y ago
> even remotely economically feasible Depends on what exactly are you mining and whether it is possible to make fuel out there. And with launch cost falling drastically I expect to see this in my lifetime.
131.
by terafo 4y ago
SpaceX is something that Musk created and controlled fully for the entirety the full period of existence of SpaceX. I think it is delusional to think it is successful despite Musk, especially with how many better funded space players such a
132.
by terafo 4y ago
I don't see them making any kind of comeback any time soon, since they can't get any high-tech equipment from abroad. Which is pretty much impossible with current leadership and current political system.
133.
by terafo 4y ago
They are sold as devkits any you need to install software on them on your own. And, IIRC, their software is meant to be used as drive-assist and doesn't work if user does not look at the road.
134.
by terafo 4y ago
$5.57M last year, probably more this year considering how hard it was to get chips last year.
135.
by terafo 4y ago
That's what he is talking about. There are different standards for safety for experimental planes and commercial jets. And he is much better in building former than latter.
136.
by terafo 4y ago
How compatible is it with modern distros?
137.
by terafo 4y ago
I spent around 45 hours there in my first walkthrough, can imagine how did you beat it so fast, but can't understand why.
138.
by terafo 4y ago
Epic owns RAD, so it won't be a problem.
139.
by terafo 4y ago
You are underestimating cost of rendering and overestimating cost of machine learning models. And forgetting the fact that cameras are already used in modern headsets for position tracking and more(there is Quest 2 demo where you cast spell
140.
by terafo 4y ago
You are hilariously wrong if you think it will not improve. Combination of prompt engineering, better and bigger datasets, better architectures, bigger models, artifact-fixing tools and sheer human creativity will seriously challenge tradit
141.
by terafo 4y ago
I wonder how expensive they are to produce now and have they managed to break even. Gwen Shotwell told the press that their costs were around 1300 with some updated model that was released in early 2021, she said that in April. Since then t
142.
by terafo 4y ago
No, they aren't even on the same architectures now.
143.
by terafo 4y ago
Turing is an evolution of Volta, but they are different architectures. A100 and GA102 DO NOT have same core architecture. 192KB of L1 cache in A100 SM, 128KB in GA102 SM. That already means that it is not the same SM. And there are other di
144.
by terafo 4y ago
Nvidia have been making different architecture for gaming and datacenter for few generations now. Volta and Turing, Ampere and Ampere(called the same, different architectures on different node). And Hopper with Lovelace are different archit
145.
by terafo 4y ago
It was on 16GB model. Can you provide details how do you run them and how fast it can run.
146.
by terafo 4y ago
It is VASTLY slower than 2080Ti. There is no competition at all. I don't know exact performance numbers, but I would estimate that M1 takes more than a minute to generate one image. UPD: asked a friend to generate image on m1 with diff
147.
by terafo 4y ago
I think you should wait until late November for dust to settle(budget segment won't change that much because of it if you don't consider buying used). By then both Nvidia and AMD will release their cards(it is likely that Intel to
148.
by terafo 4y ago
It is hard to tell since you don't provide exact numbers. I've seen SD generating 50 samples 512x512 image on 2080 Ti in 8 seconds, but it is almost certain that there is a faster version out there.
149.
by terafo 4y ago
But 4090 isn't the Titan. It doesn't have Titan drivers and everything.
150.
by terafo 4y ago
I'd guess it got something to do with introduction of Jax.

More ›