How does this compare to xarray, which also introduces named axes
on numpy arrays?
2,577 karma · joined December 29, 2019
For transposes, np.einsum can be easier to read as it let's you use (single character, admittedly) axis specifiers to "name" them.
> The basic styles of Kermit (Regular, Bold, Italic, and Bold Italic) are available today in Office, with the remaining 38 styles arriving in early May.
...from the last paragraph of the linked article.
You could argue that the attention part of the network is some form of truth validation of the next predicted token? But indeed, the current chat interfaces don't change their previous text output retroactively.
Still, I am not really convinced. If we assume a human can reason, what does "evaluation of the truth value" mean? Thinking about something? This is still performed with our implicit mental model of the world, coming from shadows on a cave's wall, right?
I'm not saying that's incorrect, but thb that's exactly the tautology I was talking about!
I'm still not on board with the (seemingly prevalent) notion that LLM's can't reason. What's reasoning, anyway? I'm not actively advocating for any side, but the arguments against reasoning always felt very tautological to me.
Fyi, the perk to self-destruct the last block didn't seem to work.
It's focused on indexing and lacks the batching idea, though. Combining named axes and vmap is a great move!
But tbh I'm turned off by the repo not being a Python package more than I'd like to admit. Reading
> Download it and put it in your working directory.
when skimming a project makes me think it's immature. Which definitely is not the case here!
But what matters is the software you're using, and that will be modern software (mostly). Modern browsers visiting JS-heavy websites, or a heavier office suite than 15 years ago.
And if this is the context, then computers must get slower over time, right?
Maybe these are "logarithmic" beings, as opposed to us "linear" beings? The lemniscate is based on geometric mean, which is basically multiplicative mean and/or mean in log-space -- as opposed to the additive mean in linear space.
If we assume we are linear beings good at intuitive addition but somewhat bad at intuitive multiplication, there could exist beings which live in log-space and whose minds are based on multiplication. Their circle would be the lemniscate.
I know that there's strong prior knowledge but the pixelation is destructive so the problem is very ill-posed.
I keep wondering whether this browser interface would be possible on a static website with all code running client-side.
How would this be possible "without the need for an account or any server infrastructure" claimed by this project?
It's the only textbook I really read during my university education (sad to admit). Even though I consider it a textbook and not pop-sci, it's incredibly approachable and teaches a framework about dynamical systems which was completely missing from all other courses I took.
Sounds like a humble brag. But it being the only book I managed to read really makes it stick out for me!
https://borgbackup.readthedocs.io/en/stable/faq.html#can-i-c...
It is not "heavily" discouraged. But you have to pay extra attention to perform a clean copy of the borg repo's files, and ideally check both instances regularly for integrity. I would assume it's easy to forget validating the "cold storage" copy of your borg repo in practice.
For non-Germans: it's a (now colloquial) term for stealing in low quantities and out of direct necessity for your own/your family's survival. In that it alleviates the "base motives" part of the crime.