tubby12345

16 ms·

91.
by tubby12345 5y ago
i interviewed at DE Shaw not less than two weeks ago. my research area is GPUs and etc. they don't have a GPUs team. i passed on the offer.
92.
by tubby12345 5y ago
http://isomorphisms.sdf.org/maxdama.pdf
93.
by tubby12345 5y ago
>topological quantum field theory lolol. i will bet all of my 401k that they absolutely do not. i've interviewed at a couple of hedge funds and have lots of friends at various prop shops/hft firms/etc. most of them are usi
94.
by tubby12345 5y ago
I believe you but I'm wondering what that memory scheduling actually looks like (e.g. solving the full MIP?). do you have a code pointer to that scheduling routine?
95.
by tubby12345 5y ago
>The memory usage and schedule of a given program/graph is statically optimized by the compiler Is it really though? The only thing I see is https://github.com/tensorflow/tensorflow/blob/95cdeaa8c848fd
96.
by tubby12345 5y ago
Can't go far on HN these days before some hot take makes you roll your eyes. The number of Borg cells that run DM jobs is a rounding error relative to the number running TF jobs.
97.
by tubby12345 5y ago
My implementation is multithreaded and parallel (ie I have multiple cores). The way I handle this is to have the leader check if it's still the leader before it does anything. That flag (`iamtheleader`) is set and read behind a mutex.
98.
by tubby12345 5y ago
Fun fact: in the extremely highly cited "In Search of an Understandable Consensus Algorithm (Extended Version)" there's a gap. In the otherwise very good figure 2: >If successful: update `nextIndex` and `matchIndex` for fo
99.
by tubby12345 5y ago
>But now, minecraft also has a functor F:(openM)→E, where E is the category of entities, which has as objects just sets of entities and as morphisms the surjective maps I too learned about categories, functors, and morphisms in my set th
100.
by tubby12345 5y ago
And you completely misunderstood the intention of my response: people should try to look beyond their own fragile sensibilities the contribution rather than the aesthetics. Barring that they should just pass (instead of criticizing).
101.
by tubby12345 5y ago
People are so entitled that they demand 1000s of hours of free (OSS) dev and aesthetics that please them. If the name is off-putting to you, you're free not to use it. If lots of people agree with you (that the off-putting name is more
102.
by tubby12345 5y ago
>And this is where I stopped God I'm so sick of these holier than thou Richard Stallman level unrealistic dismissals. This take isn't brilliant - we all, at all times, know that we have the option to not ____ if we so choose. N
103.
by tubby12345 5y ago
>who wants to devote a serious portion of their life to studying ... She's teaching classes that are being attended by students. Doesn't that mean she has a product that has some demand? Maybe she should be better compensated?
104.
by tubby12345 5y ago
I am 100-.9^100 percent sure you are just using this as an opportunity to shill for your thing but the Meta logo is just an undulating circle, while yours is a Mobius strip (meta is only the Mobius in profile)
105.
by tubby12345 5y ago
>But that's exactly my point. Zuckerberg isn't starting fresh every year. every morningl m-f, the markets open and FB starts fresh like every other ticker. every evening FB is either up or down. enough down days and zuck would
106.
by tubby12345 5y ago
>And I'm hardly alone here. Indeed, the whole reason Facebook is in public doghouse right now is Zuckerberg's focus on dominance and profit without regard to little externalities like genocide. So you may have different goalpos
107.
by tubby12345 5y ago
>I don't think failing to innovate and then buying your way out of the hole is a sign of particular acumen. I mean this is classic moving of goals posts - the only measure of acumen the CEO of a publically traded company is that lit
108.
by tubby12345 5y ago
indeed small resolutions (64x64) but i mean 2x speed is still nothing to sneeze at.
109.
by tubby12345 5y ago
the model that i got 20x on is very simple - just a couple of convs and relus - it's for edge detection on a pseudo-embedded platform (jetson) - but the wins from cuda graphs are from two things: complete elimination of kernel individu
110.
by tubby12345 5y ago
you know what i still don't understand? what's taking so long to warm up? i see that there are graph passes that run to do various things at the TS IR level, but I don't see any stats being collected (on shapes) or something
111.
by tubby12345 5y ago
There's another level of speed you can unlock by combining with https://pytorch.org/docs/master/notes/cuda.html#cuda-graphs . i got (i kid you not) 20x speed on batch size = 1 inference by first using ten
112.
by tubby12345 5y ago
Having gone through the process and coaching a friend though it, it's as much of a crapshoot as studying for any standardized test (which is to say not at all). The things on the test are standard grad school ML stuff with a dash of sy
113.
by tubby12345 5y ago
I live two blocks off Michigan by the Hancock. I can look out my window and see the cars on Michigan. There are exactly zero taxis right now, 9AM Thursday morning. I have no clue what you're talking about when you say you can hail a ca
114.
by tubby12345 5y ago
>Starting today as a PhD student in deep learning is career suicide lol maybe if you want to go into academia but, speaking from experience, FAANG is paying handsomely for this skillset.
115.
by tubby12345 5y ago
Goodfellow's is not good. Hodgepodge of material with scattered theorems and way too much attention paid to GANs (surprise surprise). ESL is very good but at this point it's the TAOCP of ML book recommendations (besides have ste
116.
by tubby12345 5y ago
Why exactly would you study math from 65 years ago that isn't relevant today as determined by the community itself ?
117.
by tubby12345 5y ago
>they were probably not very complicated They use linear models and soft cores. Plenty complicated for the kind of arb you could get executing 500ns tick to trade better than the next firm.
118.
by tubby12345 5y ago
every FAANG is almost 20 or over 20 years old. Certainly engineers that have spent those last 20 years (if they haven't already) will retire probably over the figure in the spreadsheet. Will the particular FAANG companies last another
119.
by tubby12345 5y ago
Oh gotcha. I thought you meant that you were some how modeling the prng using some reduced set of constraints.
120.
by tubby12345 5y ago
How do you pick how many constants and they're relationships?

More ›