nerdponx

41 ms·

781.
by nerdponx 3y ago
What's weird is that I don't think I've ever heard any serious practitioner doubt its capabilities . The main complaints I hear are just annoyance with pyplot.
782.
by nerdponx 3y ago
I recently started using Holoviews/Hvplot/Geoviews/Datashader for some big-ish data viz, and it's been mostly positive, but with some notable annoyances. The performance on big data is excellent. But I spent maybe two ho
783.
by nerdponx 3y ago
The object-oriented API and its documentation have improved markedly over the years, I encourage you to give it a try. Most of the official documentation examples use it now. Yes, there are lots of examples out there using the pyplot interf
784.
by nerdponx 3y ago
You're tilting at windmills here. Where in this thread do you see anyone taking about the LLM as anything other than a next-token prediction model? Literally all of the pushback you're getting is because you're trivializing t
785.
by nerdponx 3y ago
It's reductive because it obscures just how complicated that `P(next_word|previous_words)` is, and it obscures the fact that "previous_words" is itself a carefully-constructed (tokenized & vectorized) representation of a
786.
by nerdponx 3y ago
Not really, and no. Torch and CUDA align computation to hardware. If it were just a matter of doing that, we would be fine with fully-connected MLP. And maybe that would work with orders of magnitude more data and compute than we currently
787.
by nerdponx 3y ago
I think you're downplaying the importance of the attention/transformer architecture here. If it was "just" a matter of throwing compute at probabilities, then we wouldn't need any special architecture at all. P(next
788.
by nerdponx 3y ago
That's one way to think about it. It's clever way to encode "position in sequence" as some kind of smooth signal that can be added to each input vector. You might appreciate this detailed explanation: https://
789.
by nerdponx 3y ago
> TBF there is no good explanation why it works My mental justification for attention has always been that the output of the transformer is a sequence of new token vectors such that each individual output token vector incorporates contex
790.
by nerdponx 3y ago
> see the story from a few days ago I missed it, can you add a link or at least the post title I can search for?
791.
by nerdponx 3y ago
I am a perpetual newbie when it comes to things like this. What's the advantage of dlopen()-ing instead of dynamic linking?
792.
by nerdponx 3y ago
Still trying to get our team Copilot subscription for 2 months now.
793.
by nerdponx 3y ago
What is pygann? It's not on PyPI.
794.
by nerdponx 3y ago
Even without an IDE, I use Mypy like a test suite. It catches real bugs that would be either hard to find in testing, or intrusive and annoying to test for.
795.
by nerdponx 3y ago
That's always been part of the fun.
796.
by nerdponx 3y ago
It's surprising, but in a weird way it makes sense if you think about it as "learning a pattern" instead of "doing math". The model doesn't have to understand logic as we know it (and this class of models doesn
797.
by nerdponx 3y ago
> A unanimous three-judge panel of the New Orleans-based 5th U.S. Circuit Court of Appeals on Thursday agreed with Inhance Technologies that the EPA overstepped its authority by issuing the orders, since they were rooted in a section of
798.
by nerdponx 3y ago
"Empty your cup", as they say.
799.
by nerdponx 3y ago
Plenty of kids in high school were BBM/texting during class on their flip phones, Sidekicks, etc. They were told not to do it, but they did it anyway under the desk. Only a handful of strong-willed teachers actually took the phones awa
800.
by nerdponx 3y ago
> It's wild to know the movie Apollo 13 is closer now to the events than to the present. In hindsight, it seems like the middle of the 20th century was right in the middle of an "S-curve" for societal change and technology
801.
by nerdponx 3y ago
As long as they figure out the prion safety issue, "going home" and then being turned into food doesn't sound like such a bad system really.
802.
by nerdponx 3y ago
"You gotta tell people... The oceans are dying, plankton's dying... it's people. Soylent Green is made out of people."
803.
by nerdponx 3y ago
Because we consume and interact with so much of our information via browser that it's a natural pairing? Why does my browser have a PDF reader and print manager built in?
804.
by nerdponx 3y ago
Is there a Tree Sitter grammar for AWK? That could be a good way to go, via the Python bindings for Tree Sitter.
805.
by nerdponx 3y ago
What I don't understand is, where are all the code and security contributions from Big N and other multi-billion-dollar international scale and users? Do they all have their own internal fork of every major library? If not, you would t
806.
by nerdponx 3y ago
How about this? alias git='TZ=UTC0 git'
807.
by nerdponx 3y ago
Right. It might be more interesting to consider the count of Ts and Hs instead of considering exact sequences.
808.
by nerdponx 3y ago
Isn't this the reason why AGPL has started to get more popular? Everyone has to play by the very strict rules except the copyright holder, who can do whatever they want, but the community still benefits from the core software being o
809.
by nerdponx 3y ago
> Any article that begins with "using ... models" can safely be disregarded. Disregard all of biology, physics, and engineering. Got it.
810.
by nerdponx 3y ago
Wikipedia tells me that this is also the last major depiction of her death as death, rather than "assumption", which I find kind of interesting.

More ›