runeblaze

9 ms·

91.
by runeblaze 2y ago
In addition to data, having the infra to scale up training robustly is very very non-trivial.
92.
by runeblaze 2y ago
Did not read too deep into the original post, but if you use arrow you can (not sure if one ever should) do random lookups into the storage buffer and wipe out bytes + put tombstones, at least if the column is encoded "naively".
93.
by runeblaze 2y ago
Half jokingly if `s, t, a, b` ever becomes universal then either we all ascended to purity heaven or a mass extinction on the scale of K-Pg extinction occurred and wiped out most of earth's programming languages.
94.
by runeblaze 2y ago
It is a fine argument, but I mean it is youtube and it is kids we are talking about. It's really hard to show kids kindness through free content if you want to be nuanced.
95.
by runeblaze 2y ago
I know what you mean, but MrBeast cured 1000 people of (a form of) blindness, which is quite a benefit to humanity [1]. I would not be surprised if kids learn a bit of "kindness is good" from him. [1]: Of course among other things
96.
by runeblaze 2y ago
It's not that weird -- your next AA game or AAA game will probably not use it, but in React (reducer, or Redux), very interactive stuff, immutability is often seen. Also Clojure has better performance prospects than Ruby and Python,
97.
by runeblaze 2y ago
This is a bit confusing to me because I am quite sure nearly everyone <10 YoE needs guidance one way or another. Similarly almost every dev needs real code review. Without code review everyone breaks things all the time.
98.
by runeblaze 2y ago
This is oversimplifying too much IMO. Obviously the potential reward working at startups is much higher than megacorps, and you can very safely say that people working at startups have a higher risk appetite. However, plenty of people work
99.
by runeblaze 2y ago
I think a second reason might be that translating OCaml/Haskell concepts to Python has that academic connotation to it. Rust also (thanks to PyO3) has more affinity to Python than the ML languages. I guess it isn't a surprise that
100.
by runeblaze 2y ago
Probably not Ebiten for 3d games. To be fair at this point when you are doing somewhat specialist things Go starts to lose its edge. I remember trying to replicate some Numpy code in Go and that was a pain. However that's just because
101.
by runeblaze 2y ago
My uninformed guess is that at least it will cost you the branching because you need to check if it is inlined or not, and you pay that for every string. Branch prediction is likely very good for this case though.
102.
by runeblaze 2y ago
I second the conclusion as (a brutal conclusion, but still) to stop using Haskell. Haskell allows imperative-like code but the ergonomics for day-to-day big-tech engineering is far from good. The state monad or lens are excellent tools to r
103.
by runeblaze 2y ago
Since we are onto game recommendations I recommend Pokemon without any irony. The wiki is vast. Honestly I don't know how people are supposed to learn about all the deeper game mechanics without a wiki, and RNG manipulation can be quit
104.
by runeblaze 2y ago
Sorry it was a bit half joking. I am an indie game dev but at the same time I also had prior experience doing ML engineering. FWIW some types of game devs are actually versatile people for ML engineering. They can write low level code; they
105.
by runeblaze 2y ago
It was a research scientist role. The company I will omit because frankly I heard multiple companies doing it (to the point that at least two sources recommended to me to read minGPT before interviewing, so it is "invert binary tree&qu
106.
by runeblaze 2y ago
Unfortunately no "write me multi-head self attention by hand" :(, so not good enough for 2024's DL interviews.
107.
by runeblaze 2y ago
Another answer is that 3SAT and co can be seen as distilled variants of proving statements. Well, 3SAT is famously hard.
108.
by runeblaze 2y ago
I do agree that it is a code smell. However given that this is an algorithms article I don't think it is exactly that fair to judge it based on code quality. I think of it as: instead of writing it in pseudocode the author chose a real
109.
by runeblaze 2y ago
I used to code in Ruby for indie games (well, RPG Maker). For me the reason is several things: 1. Its performance meant that it would never survive against Java/C# for AA/AAA games, not to speak of C++ 2. Lack of libraries. Last t
110.
by runeblaze 2y ago
I am non-native speaker and can confirm that I can't remember how many times my grammar checker complains about my uses of "softwares" and "taxons"
111.
Show HN: Python Library to "Shuffle" ML Data Instantly(github.com)
1 points by runeblaze 2y ago|0 comments
112.
by runeblaze 2y ago
That page is written for marketing. My experience with industry research labs is that just like many places with somewhat misplaced incentive structures, a lot of the good research does not get marketed properly. e.g., the research marketin
113.
by runeblaze 2y ago
̶I̶I̶R̶C̶ ̶g̶l̶o̶o̶ ̶i̶s̶ ̶C̶P̶U̶ ̶t̶e̶n̶s̶o̶r̶s̶ ̶o̶n̶l̶y̶ ̶s̶o̶ ̶l̶i̶k̶e̶l̶y̶ ̶n̶o̶t̶ Edit: I had a brain freeze or something... gloo is not CPU only but for whatever reason I don't see it outside of CPU-comms
114.
by runeblaze 2y ago
Reading papers is absolutely part of academic research. The other part is probably producing things and submitting to journals/conferences advancing things based on what one has read (and hopefully having a day job that pays for the pu
115.
by runeblaze 2y ago
It is useful but at the same time different teams IIRC had different design doc structures. I was overwhelmed by go/greendoc when I was an intern and I just opened an empty doc and started writing based on my team's existing desig
116.
by runeblaze 3y ago
Well Peter Hessler writes like this. By the way we all know American higher education is often interspersed around the US, so Pittsburgh is actually a nice median -- not a college town but far from a big city like Chicago/NYC. The ment
117.
by runeblaze 3y ago
Ideally we can make progress on this. I saw people occasionally do weird things to bypass this, like `let x = 1 in f(x)` manually transpiled to `(lambda x: f(x))(1)` or thank god we have walrus now, so `(x := 1, f(x))` (as mentioned in the
118.
by runeblaze 3y ago
Just want to put out here that being neurodivergent is an easy route to lose hope. Most people / professors are neurotypical and they guide you through their neurotypical lens, as they have previously found success doing.
119.
by runeblaze 3y ago
Btw, totally unsolicited advice, but please take care of yourself. You sound like you have quite some decisions to dwell on, and correct decisions are much harder in moments of hopelessness. If you feel hopeless, we (at least me, and HN in
120.
by runeblaze 3y ago
Good point, and my original opinion should at least have what you said as fine print. Personally, I am finishing up a PhD in CS. When I started my PhD, deep learning had already been all the craze but I chose another subfield -- that did no

More ›