gyrovagueGeist

10 ms·

31.
by gyrovagueGeist 3y ago
I've been convinced, thanks!
32.
by gyrovagueGeist 3y ago
While "local maximas" is wrong, I think "a local maxima" is a valid way to say "a member of the set of local maxima" regardless of the number of elements in the set. It could even be a singleton.
33.
by gyrovagueGeist 3y ago
Yep, it went from compelling problems to everyone is an idiot to drive drama real fast.
34.
by gyrovagueGeist 3y ago
Another relevant paper is: https://epubs.siam.org/doi/10.1137/15M1032168
35.
by gyrovagueGeist 3y ago
The steve jobs example has a typo, "place of dearth" ( https://kagifeedback.org/assets/files/2024-03-05/1709652268-... )
36.
by gyrovagueGeist 3y ago
All the more reason to build a proper moon base with an indoor court
37.
by gyrovagueGeist 3y ago
> I had recently read The Phantom Tollbooth, where a kid in the real world crosses a magic threshold and enters a cartoon world. This felt like that. I wanted more. This feels like a weird lead in to me. It invites comparisons to the sto
38.
by gyrovagueGeist 3y ago
Well... academic code is even worse as it has no incentives other than those (user bases are ~1-2 people and the paper is the product). As long as it worked briefly to produce the paper results it's "gtg".
39.
by gyrovagueGeist 3y ago
Yep, the language is consistent! Although you can usually always get by with the later.
40.
by gyrovagueGeist 3y ago
It doesn't make sense to me either. Universities barely even fund the lifecycles of their own research based software projects.
41.
by gyrovagueGeist 3y ago
This is the goal of the RSE field, but it's often still quite rare :( https://us-rse.org/
42.
by gyrovagueGeist 3y ago
Taking the example in this comment thread, even if the model takes an arbitrary nonparametric distribution of input temperatures and perfectly returns the posterior distribution of freezing events there is still a difference in model error
43.
by gyrovagueGeist 3y ago
They can also mean pushed forward uncertainty from input parameters which isn't exactly the same as model error
44.
by gyrovagueGeist 3y ago
The frequency is very different
45.
by gyrovagueGeist 3y ago
FYI, the fraction, alignment, and exponent syntax is not correct on this output and it won’t render the same.
46.
by gyrovagueGeist 3y ago
Even OpenBLAS (the default iiuc) does all of that and more to optimize for different levels of the cache hierarchy: https://www.cs.utexas.edu/~flame/pubs/GotoTOMS_revision.pdf I'm not sure where/how they
47.
by gyrovagueGeist 3y ago
well kinda... Numpy is rather slow relative to what you can build with a pure C/C++/Fortran/Rust pipeline. Its API usually prevents memory reuse, putting memory allocations on the critical path. You can't fuse operators.
48.
by gyrovagueGeist 3y ago
What is your BLAS backend?
49.
by gyrovagueGeist 3y ago
Still no stochastic rounding :(
50.
by gyrovagueGeist 3y ago
Taichi solves a different problem. It's (mainly) a tensor compiler that targets kernel generation on different architectures, while Chapel is for orchestrating and communicating between computations over distributed memory spaces (and
51.
by gyrovagueGeist 3y ago
It's an iterative thing! Neural Network based surrogate models are becoming popular, but you need the compute to solve the PDEs to train the AI and compute the loss in the first place ;)
52.
by gyrovagueGeist 3y ago
oh no! I just read Eifelheim 2 weeks ago, it was a lovely book!
53.
by gyrovagueGeist 3y ago
Very nice! I think the landing README should really be a comparison with CuPy and not NumPy though
54.
by gyrovagueGeist 3y ago
Kagi as been taking 2+ seconds over the past week and has felt at bit laggy for me (currently on the trial) It was very fast earlier this month though!
55.
by gyrovagueGeist 3y ago
Although it is not as hyped, HPC is much wider than just machine learning. H100 are incredibly useful in numerical simulation and scientific computing.
56.
by gyrovagueGeist 3y ago
Based on the name of the post I assume you've already looked/are aware of it, but just it case, check out some projects from past Data Science for Social Good summer programs/fellowships?
57.
by gyrovagueGeist 3y ago
> But more importantly, how does it differentiate between "Your delivery has arrived" and "this weekend, 10% off all deliveries paid with mastercard from chase banks" IIUC it is up to the app developer to tag the noti
58.
by gyrovagueGeist 3y ago
A sufficiently complicated F can include an intermediary calculation of G for future token steps. This is not explicitly modeled or enforced for LLMs (and doing so would be interesting) but I'm not sure I could say with any sort of con
59.
by gyrovagueGeist 3y ago
I'm surprised it's not listed on the README directly but in case you looked for it like me :) here is what their site says about the alternative you're probably thinking of: "toss cleverly encodes port information in the
60.
by gyrovagueGeist 3y ago
It has nearly the same interface design as cuFFT, admittedly with less documentation. I’m not sure I understand this complaint versus the competition. https://rocm.docs.amd.com/projects/hipFFT/en/latest/a

More ›