notsylver

6 ms·

31.
by notsylver 2y ago
I've been digitising family photos using this. I scanned the photo itself and the text on it, then passed that to an LLM for OCR and used tools to get the caption verbatim, the location mentioned and the date in a standard format. That
32.
by notsylver 2y ago
Could you share your ACL setup? I haven't had time to look at it much but this sounds like exactly what I want to do.
33.
by notsylver 2y ago
opt-in but afaik they still show up in places unless you disable them
34.
by notsylver 2y ago
I'm stuck using somewhat unreliable starlink to a datacenter ~90ms away, but I can run 7b models fine locally. I agree though, cloud completions aren't unusably slow/unreliable for me, it's mostly about privacy and it be
35.
by notsylver 2y ago
This looks a lot more impressive than a lot of GitHub Copilot alternatives I've seen. I wonder how hard it would be to port this to vscode - using remote models for inline completion always seemed wrong to me, especially with server la
36.
by notsylver 2y ago
I think it would be more interesting doing this with smaller models (33b-70b) and see if you could get 5-10 tokens/sec on a budget. I've desperately wanted something locally thats around the same level of 4o, but I'm not in a
37.
by notsylver 2y ago
Could you get the speed and location of the plane and estimate how far it should move across two images to determine if its the right target?
38.
by notsylver 2y ago
I've had the opposite experience - I tried continue.dev and for me it doesn't come close to copilot. Especially with copilot chat having o1-preview and Sonnet 3.5 for so cheap that I single handedly might bankrupt microsoft (we ca
39.
by notsylver 2y ago
I doubt it, but it would be interesting if they recorded Stadia sessions and trained on that data (... somehow removing the hud?), seems like it would be the easiest way for them to get the data for this.
40.
by notsylver 2y ago
i've been using it for... years? and it still feels like magic when i use it. v4 fixes my only real issue with it which was the config file feeling so detached and rough compared to everything else.
41.
by notsylver 2y ago
I found 4o to be one of the worst, but I was using the API. I didn't test it but sometimes it feels like images uploaded through ChatGPT work better than ones through the API. I was using Gemini Flash in the end, it seemed better than
42.
by notsylver 2y ago
That's interesting. I downscaled the images to something like 800px but that was mostly to try improve upload times. I wonder if downscaling further and with a better algorithm would help.. I remember using CLIP and found different sca
43.
by notsylver 2y ago
I tried open source and closed source OCR models, all were pretty bad. Google vision was probably the best of the "OCR" models, but it liked adding spaces between characters and had other issues I've forgotten. It was bad eno
44.
by notsylver 2y ago
I've been doing a lot of OCR recently, mostly digitising text from family photos. Normal OCR models are terrible at it, LLMs do far better. Gemini Flash came out on top from the models I tested and it wasn't even close. It still h
45.
by notsylver 2y ago
yeah, ive used llama.cpp grammars before, which is why i was thinking about it. i just think it'd be cool for llamafile to do basically that, but with included defaults so you could eg, require JSON output. it could be cool for prototy
46.
by notsylver 2y ago
As soon as I saw this was part of llamafile I was hoping that it would be used to limit LLM output to always be "valid" code as soon as it saw the backticks, but I suppose most LLMs don't have problems with that anyway. And I
47.
by notsylver 2y ago
Solved, just not by Tesla who have been promising it every year since.. 2016?
48.
by notsylver 2y ago
I was planning to do this myself lol. I was going to use SQLite as the index, and use `sqlite-vec` or something similar to query for similar files directly. I think the only other thing I was planning were more filters, `"positive term
49.
by notsylver 2y ago
It might be because I'm not used to SQL, but I've found the opposite. Writing a large query with lots of conditions (eg, if the user is signed in, hiding content they've blocked) is miserable without an ORM that can build the
50.
by notsylver 2y ago
But then no one bothers implementing your document format because their existing HTML already works and it's not worth investing more time into such a small userbase, and then you're maintaining even more code
51.
by notsylver 2y ago
I hope storage boxes come to Singapore, they're ridiculously affordable for backups. I don't see it happening though, considering the US datacenter still doesn't have them available. I can dream I guess - the ping from AU to
52.
by notsylver 2y ago
Isn't that what ChromeOS is?
53.
by notsylver 3y ago
They mention few-shot prompting in the prompt engineering section, which I think is what you mean.
54.
by notsylver 3y ago
What do you use for tab-autocomplete in VS Code? I've been trying to find something that can replace copilot, just because it sounds fun. Everyhing I've found seems more aimed at entering a prompt and having it refactor code, not
55.
by notsylver 3y ago
That shouldn't stop them from encrypting the contents and making it intentionally difficult to draw conclusions from?
56.
by notsylver 3y ago
https://githubnext.com/projects/code-brushes/ You select some code and apply a brush and it'll use a preset prompt to modify it: "Make Readable", "Add Types", "Document" are some
57.
by notsylver 3y ago
marver https://github.com/sylv/marver My NAS has a load of files dumped onto it over the years. Instead of spending a couple weekends sorting through it like a sane person, I'm making something to automatically in
58.
by notsylver 3y ago
These aren't just annoyances, they are core features of the sites.
59.
by notsylver 3y ago
IMO, the main reasons are (but are definitely not limited to): - You can fine tune these models for very specific tasks, which GPT-4 might not be as good at. - Open source models are free. You can use them as much as you want without worryi
60.
by notsylver 3y ago
I've always wondered, couldn't screenreaders and similar just disable autofocus? I can't imagine a scenario where a site would break by disabling it, all I can think of is they don't have that much control over the brows

More ›