nrub

38 ms·

1.
by nrub 12d ago
This is exactly what Taalas has already done, and the reason they got quickly acquired by AMD. Their chip runs Llama 3.1 8B at 17k tok/s, and even if llama 3.1 is dated, I can think of many problems I could use it for, especially at th
2.
by nrub 1mo ago
Right, but the point is that you can't conclude that a model is necessarily bad because it's not hitting the same scores on benchmarks. I just don't agree with lacker's conclusion, because their logic doesn't seem t
3.
by nrub 1mo ago
Or they spent time optimizing their model to real world problems they're facing and didn't waste time trying to game a benchmark.
4.
by nrub 2mo ago
That's true, but that's just one model that has some easier setup in the ecosystem. It's not easy to get a consumer gpu that has 128Gb of vram, so while the Halo chip isn't as performant, the ceiling for running larger m
5.
by nrub 2mo ago
In opencode you can get granular, allow `git log *`, deny `git add *` and `git commit *`
6.
by nrub 2mo ago
> simple chart specs can be reliable, but generated charts are often of low quality due to reliance on system defaults; - complex chart specs with explicit details can produce good-looking charts, but they are verbose and agents can stru
7.
by nrub 4mo ago
Just because they don't know how to use a computer well, doesn't mean you can't teach them.
8.
by nrub 4mo ago
I no longer keep my dotenv files in plaintext. I use `sops` to keep an encrypted env around and you can use tools like direnv to make them available to your shell while you're working. Obviously the LLM could print any of these secrets
9.
by nrub 4mo ago
No, they're not there. Did you read my comment? It's not just flawed or limited implementation, those are things that python just doesn't have.
10.
by nrub 4mo ago
Yeah, that's cool, but it would be almost completely unnecessary if python just had actual private methods/classes/properties. It's a lot like pydantic, which is completely unnecessary if you had strong typing.
11.
by nrub 4mo ago
Yes, but: a) they're a second class citizen, not guaranteed to be used in whatever niche of the python ecosystem you find yourself in and there's already an n+1 problem with multiple type checker written by third parties, rather t
12.
by nrub 4mo ago
You joking? - strong typing - real concurrency (heaven forbid you want a background task without having to spool up an external message queue and worker) - immutability - limitations in error handling (sort of just typing really) - limitati
13.
by nrub 5mo ago
You have literal Stockholm syndrome. Not joking.
14.
by nrub 2y ago
RCA with FM radio, https://en.wikipedia.org/wiki/Edwin_Howard_Armstrong
15.
by nrub 3y ago
I've seen a few suspect benchmarks for recent announcements of LLM releases. I'm sure they made an attempt at an honest benchmark, but until there's an independent assessment and benchmark (preferably multiple) you have to as
16.
by nrub 3y ago
There are promising methods developing for Physic's informed neural networks. Mathematical models can be integrated into the architecture of neural networks such that the parameters of the designed mathematical models can be learned. E
17.
by nrub 3y ago
Do what you want with some of those tips, maybe not necessarily good advice. But if you're searching, try these search queries replacing `%s` with your desired role(s): ``` search_strings = [ 'site:boards.greenhouse.io %s&#x27
18.
by nrub 3y ago
Yup
19.
by nrub 5y ago
One option if you're on github, https://github.com/marketplace/actions/todo-to-issue
20.
by nrub 8y ago
The paper linked above does directly address the case of multiple experiments occurring in the same context. They address this with hill-climbing over those 180 different variations. The use of a bayesian linear regression takes place of th
21.
by nrub 9y ago
They have some hard comparisons in some of their earlier blog posts on how spacy compares to the other popular open source NLP libraries. In my experience it has been much easier to use and faster than things like Stanford's library or
22.
by nrub 10y ago
Have you used paredit-mode in emacs with a lisp dialect? Getting proficient with this mode can be a lot like what you describe. Paredit let's you navigate and edit the tree structure of lisp code pretty effectively. It's not inher
23.
by nrub 10y ago
Artemis Health | Product/Data/Engineering Roles | Salt Lake City, UT | ONSITE, Full Time | https://artemishealth.com We build analytics and visualization tools for self-insured companies. We're funded, have had gr
24.
by nrub 10y ago
PageRank is analogous to this type of music discovery if you created a graph based on artists who have collaborated together. It'd be about the same as having a teleportation vector that always restarts the random walk back at the arti
25.
by nrub 10y ago
Looks like it's the first suggestion in their production developer guide, https://docs.angularjs.org/guide/production
26.
by nrub 10y ago
I think part of the terms for the Alexa voice service forbid auto-listening.
27.
by nrub 11y ago
I don't know if that's what he's referring to, but I will comment that KSL is a pain. It's a bloated error filled site, and I really wish the local communities would stick to CL.
28.
by nrub 11y ago
A month of inactivity and an issue that is 12 days old, and someone is trying to fork the project under a new name? Seems kind of fickle and impatient. Gogs is a great project, and that thread doesn't seem to imply that it's dead
29.
by nrub 12y ago
Additionally by Sussman is Functional Differential Geometry. This builds on work from SICM and delves into ideas of Relativity.
30.
by nrub 12y ago
Serious question here. Is microfinance a good thing? What do we know about how it affects the borrowers? Most of the things I have read have been from a lenders perspective. About profit, and investment. Is that really the goal?

More ›