electroglyph

9 ms·

1.
by electroglyph 17d ago
good stuff, no mention of confidence tho, recommend having a look at diffusiongemma and others.
2.
by electroglyph 23d ago
Opus 5 literally benchmarks higher than Fable
3.
by electroglyph 25d ago
Good on them for the retraction. It's good to see science at work.
4.
by electroglyph 1mo ago
it's not a schooner...it's a sailboat
5.
by electroglyph 1mo ago
i would assume a model this size would require finetuning tbh. even functiongemma recommends that.
6.
by electroglyph 1mo ago
my experience is the same, but deepseek is planning on increasing prices soon, which will make it a lot less attractive
7.
by electroglyph 1mo ago
yep, agreed. little bit smarter than deepseek 4 flash, but a little bit worse at tool calls.
8.
by electroglyph 2mo ago
the cats can traverse a fence, hole or not, so not sure that's relevant?
9.
by electroglyph 2mo ago
If it makes you feel any better, my apology was genuine.
10.
by electroglyph 2mo ago
https://huggingface.co/TheDrummer/Rivermind-12B-v1
11.
by electroglyph 2mo ago
Such a weird project. Does the world really need this?
12.
by electroglyph 2mo ago
i handed out so many upvotes here, good work everybody!
13.
by electroglyph 2mo ago
i just today pulled the back off my old oneplus 8 pro and put a new battery in it after putting lineageos on there. i decided i was tired of using my locked down Samsung that's full of crap
14.
by electroglyph 2mo ago
https://www.youtube.com/watch?v=uh1o_03ilQY
15.
by electroglyph 3mo ago
ONNX has AVX512 CPU kernels too, and openvino uses ONNX internally (and ONNX supports openvino backend)
16.
by electroglyph 3mo ago
it's a pain in the ass to do properly. what we really need it something like auto-round for ONNX
17.
by electroglyph 3mo ago
ONNX is my first suggestion to people looking for speed gains on CPU
18.
by electroglyph 3mo ago
heh, i came here to say basically the same thing
19.
by electroglyph 3mo ago
existing embedding models like alibaba's modernbert tune or one of the jina v5s would probably map query to category automatically. (i.e. store embeddings of each category and calculate cosine sim for each incoming query vs. categories
20.
by electroglyph 3mo ago
heh, when Raymond Chen dunks on the MSVC team =)
21.
by electroglyph 3mo ago
it's probably a lie
22.
by electroglyph 3mo ago
deepseek 4 pro is insanely good for the price
23.
by electroglyph 3mo ago
that link actually recommends not doing it from UEFI and doing it via software
24.
by electroglyph 3mo ago
any divergence (even if the benchmark is better) from full precision is error
25.
by electroglyph 4mo ago
this is better than TFA
26.
by electroglyph 4mo ago
sometimes i wonder if the left hand knows what the right is doing. it looks like we arrested our own spy in this case: https://www.politico.com/news/2026/05/25/american-journalist...
27.
by electroglyph 4mo ago
absolutely. somebody online was wanting an LLM with Georgian language support, and that's exactly what i suggested: start digitizing Georgian text.
28.
by electroglyph 4mo ago
i'll upvote this each time it's submitted
29.
by electroglyph 4mo ago
you should be using dflash with that model, look it up
30.
by electroglyph 4mo ago
heretic maintainer: https://github.com/p-e-w/heretic the fun bits are in another branch or PRs

More ›