radq

7 ms·

1.
by radq 3mo ago
I'm disappointed with the commentary here. "GPU bubble" is an industry standard term, and literally how I would describe this to my colleagues in the industry. Look for example at the second slide here https://medi
2.
by radq 3mo ago
This is what people in the field call it. I'm sorry you're offended.
3.
by radq 3mo ago
Appreciate you saying the blog was nice. Not sure what you mean by "CODEX fingerprints", but I'll engage with the other points. We work on small models, and our customers want real-time inference on modern GPUs. The sub-title
4.
by radq 3mo ago
Thank you for the kind words. We will write and share more of these.
5.
Popping the GPU Bubble(moondream.ai)
197 points by radq 3mo ago|50 comments
6.
by radq 1y ago
The 'point' skill is trained on a ton of UI data; we've heard of a lot of people using it in combination with a bigger driver model for UI automation. We are also planning on post-training it to work end-to-end for this in an
7.
by radq 1y ago
Thanks! If you could shoot me a note at vik@m87.ai with any examples of the precision/recall issues you saw I'd appreciate it a ton.
8.
by radq 1y ago
Cool project! The codebase is simple and well documented, a good starting point for anyone interested in how to implement a high-performance inference engine. The prefix sharing is very relevant for anyone running batch inference to generat
9.
by radq 2y ago
Hello folks, I work on moondream. Posted a demo video on twitter for this release: https://x.com/vikhyatk/status/1864727630093934818 Happy to answer any questions!
10.
by radq 2y ago
Not true, H100s cost $2-3/GPU/hr on the open market.
11.
by radq 2y ago
Have you considered sponsoring an open-source project? ;)
12.
by radq 2y ago
1/3rd "activated parameters", while also requiring 2x the VRAM.
13.
by radq 2y ago
The training technique used here (fitting something similar to a NeRF to different views of the same image) is pretty similar to this paper which uses a similar technique to denoise (instead of upscale) output features: https://a
14.
Show HN: Moondream, a small vision language model that runs on 8GB of RAM(github.com)
9 points by radq 3y ago|0 comments
15.
by radq 3y ago
I'm confused - you posted in the "who wants to be hired" thread, and then got an email from this company asking if you'd be interested?
16.
by radq 3y ago
Do outlier features emerge in sub-100M parameter models? I haven't seen any research discuss it below the 124M scale (bert-base). At that scale training a model takes ~4 days on an 8xA100 node.
17.
Show HN: AlignedBot, the safest and most aligned chatbot(fsck.ai)
1 points by radq 3y ago|0 comments
18.
Comparing Queuing Strategies in Distributed Systems(fsck.ai)
2 points by radq 3y ago|0 comments
19.
by radq 3y ago
The plugin is supposed to ask for confirmation, according to OpenAI's documentation at least. > When a user asks a relevant question, the model may choose to invoke an API call from your plugin if it seems relevant; for POST request
20.
by radq 3y ago
Literally happened to me, though it was during COVID lockdown so it's possible things have gotten better.
21.
by radq 3y ago
Looks like you're right, there was a recommendation to increase it but I don't see anything about it taking effect.
22.
by radq 3y ago
I don't expect a lot of people to take this up. Express Entry was already an option for these folks and you get permanent residency from the start under that program.
23.
by radq 3y ago
As someone that's used to live in Canada and is currently in the USA, this is not true. Six month wait time to get an x-ray for a hairline fracture, people dying because of long wait times for cancer screenings etc. In the USA things a
24.
by radq 3y ago
You can switch jobs on an H1-B, and if you get fired/laid off you currently get 180 days to find a new job (though that used to be 60 days previously).
25.
by radq 3y ago
More like the engineers and scientists built it, but legal and PR think it's a bad idea to release. Which they're probably right about considering the reception Galactica got.
26.
by radq 3y ago
Andrej Karpathy's "zero to hero" series [1] was how I learned the fundamentals of this stuff. It's especially useful because he explains the why and provides intuitive explanations instead of just talking about the what
27.
by radq 3y ago
Yeah, that's pretty close. It might be more precise to say they trained one big model that includes 8 "expert networks" and a mechanism to route between them, since everything is trained together. There isn't a lot of pu
28.
by radq 3y ago
If it's similar to the switch transformer architecture [1], which I suspect it is, then the models are all trained on the same corpus and the routing model learns automatically which experts to route to. It's orthogonal to beam se
29.
by radq 3y ago
Looks like Adobe makes about 26% of their revenue from EMEA, so it's clearly worth it for them to continue operating there and deal with the regulatory overhead. But is it possible for them to have a different corporate structure that
30.
by radq 3y ago
I have some reservations about the specification - it looks like there are a lot of caveats to how it can be used. For example, it's allowed in for and for-of loops, but not in for-in loops. Also looks like destructuring will not be al

More ›