giffmana.ai
Researcher (OpenAI. Ex: DeepMind, Brain, RWTH Aachen), Gamer, Hacker, Belgian.
Anon feedback: https://admonymous.co/giffmana
š ZĆ¼rich, Suisse š http://lucasb.eyer.be
301 posts
5,783 followers
200 following
Regular Contributor
Active Commenter
comment in response to
post
index the codebase first. Then ask o1 in "codebase chat" the same questions you would like to ask the author/owner of the codebase.
Mostly useful when digging into unknown/new codebases and trying to understand them. Or asking about possible bugs :)
comment in response to
post
yes, I noticed that over Christmas break and ever since, it's just... a lot more boring here.
My first reason to open social media is to be entertained. Second reason to entertain. Maybe last third to learn something new.
Not much of any of these here. I know I know, be the change and all that.
comment in response to
post
It reeeeally depends what are loss1 and loss2, both regarding whatās standard and whatās wasted.
I honestly think you are confused, the three codes in three different posts of you mean three different things. I donāt mean it in a negative way, but clearing it up would take more time than I want :/
comment in response to
post
Yeah it seems either he had a mistake in the OP, or the subject of the discussion has drifted :)
comment in response to
post
Until we got good enough AI supported search, no, you canāt realistically expect them to find anything and everything from the past 30 years when the vocab and everything changes.
comment in response to
post
Well, no, two things:
1. In the OP indeed *both* formulations waste compute, so yeah :)
2. In 2nd post, you are not doing the same thing as in your OP! In your 2nd, you are doing good old micro batching which indeed the second way is the standard way.
So what you say keeps changing O.o
comment in response to
post
I would enjoy that meeting ;)
comment in response to
post
Yeah itās silly to expect the new generation to know everything the old generation did, doing so shows complete lack of empathy.
comment in response to
post
If the two graphs are completely disjoint, then there is no point in this. If they have some commonality (like model) then this does the common part twice.
comment in response to
post
Iām somewhat confident both of these are sins lol second one wastes a ton of compute!
comment in response to
post
Not quite because this one is not stacked, so I give it better chance to scale:
comment in response to
post
Thatās what the globe was for!
comment in response to
post
One of the physics of llm papers studied that and found you need a certain amour of repetitions of a factoid before itās memorized. Repetition can be either multi epochs or just the same fact in another document. Number of needed repeats is also related to model size.
comment in response to
post
OK OK Iāll admit it, Iām feeding off your fomo! There canāt be enough fomo! Mmmm fomo!
comment in response to
post
Yeah they compress videos to shit here, and are considering making good quality videos a paying feature.
comment in response to
post
No talk but two posters, Iām just a middle author but will try to be there (locca and NoFilter). That being said my main occupation here will be meeting many of my new colleagues.
comment in response to
post
lol exactly. I said cdg whenever possible.
That being said, our flight is operated by AirCanada, including a layover in Canada, that is a lot worse.
comment in response to
post
Oops! Thanks
comment in response to
post
This afternoon flight? Iām taking that too
comment in response to
post
Aye, finally I can without it being weird lol
comment in response to
post
comment in response to
post
Yep, and related to this: bsky.app/profile/giff...
comment in response to
post
Iām confused, why not nbviewer.com which has existed and working for a decade?
comment in response to
post
huh indeed got lost. nvm then, deleting.
comment in response to
post
All the connections you 3 mention aren't wrong, but the OP figure is about origin/influence, and I'm telling you nonlocal had epsilon influence on ViT invention or development. The arrow should come from BERT, but that's not on the pic, so next best arrow source is Transformer.
comment in response to
post
Yes. ViT is not inspired by nonlocal paper at all, and I may even have been the only person (or close to that) in the project who knew nonlocal paper.
Iām not saying this as a hater, I liked nonlocal:
comment in response to
post
Yeah but the arrow from non-local to ViT is incorrect.
comment in response to
post
I agree š¬
comment in response to
post
Bought it, will frame or something š
comment in response to
post
*As mentioned: Andreas, @asusanopinto.bsky.social and @mtschannen.bsky.social did a heroic effort here and deserve all the credit for these models and the interesting report. I merely supported, advised, and handed things over. Keep a close eye on these guys!
comment in response to
post
Thanks :)
comment in response to
post
mais bon*
comment in response to
post
Oui jāĆ©tais un peu perplexe la premiĆØre fois que jāai vu Ƨaā¦ mais moins au moins ils sont cohĆ©rents et traduisent vraiment TOUT lol
comment in response to
post
Top, merci FranƧois!
Maisā¦ Ā«Ā cadre dāapprentissageĀ Ā» cāest Ā« learning frameworkĀ Ā» ou quoi? O.o
comment in response to
post
Thanks!
comment in response to
post
Most cli tools you can keep. Install homebrew.
comment in response to
post
Thanks :)
comment in response to
post
Thanks Tobias! I owe you and Marco a lot :)
comment in response to
post
Waiting for the core first authors thread :)