pedrovhb

8 ms·

31.
by pedrovhb 3y ago
Imagine if one day we come up with a technology that allows resolving _arbitrary strings_ and doesn't cost multiple times the cost of the hosting structure itself.
32.
by pedrovhb 3y ago
That's not really the entire js of the plugin, but its schema; it's generated from the OpenAPI spec which you have to provide when writing actions/plugins. When I was looking through the requests for the available plugins the
33.
by pedrovhb 3y ago
For an actually intentional, non-cursed version of this, see the nix-shell shebang [0]: > #! /usr/bin/env nix-shell > #! nix-shell -i python3 -p python3Packages.pillow python3Packages.ansicolor > > #
34.
by pedrovhb 3y ago
Was this (from the second paragraph onwards) written by GPT-4, by any chance?
35.
by pedrovhb 3y ago
Nice! Playing around with the widget under the section "Experimenting with lines" I'm reminded of Bret Victor's Inventing on Principle talk [0] (an absolute must watch, if anyone hasn't yet). In particular, changing
36.
by pedrovhb 3y ago
That's interesting. In my experience, although synctatic errors do happen, they're not nearly as common as, say, hallucinating methods or tripping over the exact API of a library. One thought I've had but haven't experim
37.
by pedrovhb 3y ago
Ah! Is Q* a play on the A* pathfinding algorithm, but with a Q presumably for the Query vector of attention mechanisms? Clever, and actually does reveal something about the project.
38.
by pedrovhb 3y ago
Latent consistency models are a pretty radical game changer that came up recently. There are LoRAs [0] that you can just use alongside any SD or SDXL that just cut the number of inference steps you need to 2-8, rather than the usual ~25+. I
39.
by pedrovhb 3y ago
I'm not sure I understand what you mean to say. To me it's a reasonable question asking whether text to speech models can complete a missing part of some existing speech audio, or make it go on for longer, rather than only generat
40.
by pedrovhb 3y ago
Perhaps the board was not necessarily in a hurry but set it up to be swift, with Sam and even MS kept in the dark until the actual firing.
41.
by pedrovhb 3y ago
Hah, me too! I read the article title and thought "Hm, that's a bit of a weird idea, but I suppose syscalls do add up" before jumping in and realizing my curiosity wouldn't quite be satisfied in the exact way I expected
42.
by pedrovhb 3y ago
I wouldn't say that. It seems analogous to saying a program uses a simple version of a document-oriented database built in to the system when in truth it's just using dictionaries. Sure, conceptually they do kind of the same thing
43.
by pedrovhb 3y ago
I've had luck with an RX5700XT and whisper.cpp built with clblast. Works like a charm, not entirely a scarring experience getting it to work (easier than most other stuff which was surprising to me). One arcane detail is that whereas f
44.
by pedrovhb 3y ago
Just wanna say, that's pretty great API design :)
45.
by pedrovhb 3y ago
I had the original mini too. Really fun, but mine is RIP in the water under a waterfall ravine. It's hard to say whether it was a GPS bug, a gust of wind, or just my own fault, but as I was shooting a video next to some rock walls, it
46.
by pedrovhb 3y ago
Interesting that they use T5 for the distilled model. I was under the impression that encoder-decoder architectures were on the way of the Dodo, but it seems they may still be relevant after all. Also interesting is that this isn't an
47.
by pedrovhb 3y ago
I just wish xattrs were more resilient and explicit.
48.
by pedrovhb 3y ago
A great framework for doing something along those lines is Frida ( https://github.com/frida/frida ). Works on a bunch of stuff, including Android and iOS. Some global-ish certificate pinning bypasses work through Frida,
49.
by pedrovhb 3y ago
I'm not sure I understand. Isn't vector storage about storage and retrieval strictly, being a separate thing from the embedding generation itself? My mental model is that the job of vector storage is to just fetch the closest ve
50.
by pedrovhb 3y ago
That's nice! I've been thinking a lot about models running code lately, and using GPT-4 with its code interpreter to try out ideas quickly. Have you considered, as OpenAI does, using IPython for this? What strikes me is that there
51.
by pedrovhb 3y ago
Clipboard history, Pop! Shell, and workspace scroller (scroll across the top bar to quickly go through workspaces), on NixOS. I also set up dconf options so I always have 10 workspaces, not just one more than the number currently being used
52.
by pedrovhb 3y ago
Given the massive imbalance in the memory bandwidth bottleneck, I wonder why specialized hardware is the way it is. Is there some use case in which processing is the bottleneck, or at least it's more even? Are we expecting some softwar
53.
by pedrovhb 3y ago
It doesn't, you'd have to unpin certificates in whatever application you're using. Browsers generally don't do certificate pinning though, they'll (correctly) respect your installed certificates, which would include
54.
by pedrovhb 3y ago
I haven't seen it used by users, but it's amazing for non-technical staff (PMs, support to some extent, etc). They can inspect and modify the database while making sure everything still validates, run batch jobs, actions, etc and
55.
by pedrovhb 3y ago
Yup. No project of this size is without some ugliness, but Django really is an impressive piece of technology and complex but cohesive architecture. The more I understood its layers the more I appreciated how much it does, and how well it a
56.
by pedrovhb 3y ago
Not sure if that's still the case. I remember having trouble building it a couple of months ago, had to tweak the Makefile because iirc it assumed ARM64 <=> Mac, but I recently re-cloned the repo and started from scratch and it w
57.
by pedrovhb 3y ago
I've been playing with running some models on the free tier Oracle VM machines with 24GB RAM and Ampere CPU and it works pretty well with llama.cpp. It's actually surprisingly quick; speed doesn't scale too well with the nu
58.
by pedrovhb 3y ago
There's plocate too, which is a bit faster still according to the authors.
59.
by pedrovhb 3y ago
Isn't that essentially beam sampling?
60.
by pedrovhb 3y ago
Very handy, thank you!

More ›