kzrdude

7 ms·

91.
by kzrdude 2mo ago
27B means it has 27 billion parameters, but how that translates to model size depends on the model architecture. For Qwen 3.5 architecture (same as 3.8) the native size is around 55 GB memory. So you need quantization to fit it in 32 GB. Go
92.
by kzrdude 2mo ago
Funny but shows how it doesn't have a useful world model
93.
by kzrdude 2mo ago
Are they forgetting that people use Claude for presentations and business reports too?
94.
by kzrdude 2mo ago
Why is it connected to number of parameters?
95.
by kzrdude 2mo ago
This comment by the Z AI lead is relevant, about parameter vs data scaling: https://news.ycombinator.com/item?id=49357405
96.
by kzrdude 2mo ago
Mirror: https://xcancel.com/jietang/status/2089941544581403107
97.
by kzrdude 2mo ago
Is also a Swedish word, though slang, the imperative form of paxa which basically means the same as "dibs!" If I know my english colloquialisms correctly.
98.
by kzrdude 2mo ago
A lot of them are Scandi place names, too.
99.
by kzrdude 2mo ago
The German vowels are a good start, but Swedish pronunciation is of course a little bit vowel shifted from German.
100.
by kzrdude 2mo ago
Lots of used bookstores are online, and I’ve ordered lots of books that way. This website for example may look like a user-to-user book sales platform but most books there are supplied by used bookstores: https://www.bokborsen.se
101.
by kzrdude 2mo ago
Then we should change the law
102.
by kzrdude 2mo ago
That is just a policy problem. A reactive legislature could fix this in a week, rewriting the law. For example, require the digitized copies to be put in a public trust where for example libraries can provide these books as digital loans. (
103.
by kzrdude 2mo ago
We don’t know but I think we assume that it is MoE or some almost equivalent sparse activation technology.
104.
by kzrdude 2mo ago
For some tasks yes, and we don't know how many active parameters Luna is using..probably less than 27B
105.
by kzrdude 2mo ago
Raw model size is around 27x2 GB since it's in BF16 format
106.
by kzrdude 2mo ago
It's fun to see "test time" scaling work out so well, maybe the best example of all.
107.
by kzrdude 2mo ago
There is an extra green square where no egg is present, so it's a false positive in the expected.
108.
by kzrdude 2mo ago
Great, happy that it was helpful
109.
by kzrdude 2mo ago
In the third vision bench result, Sol is 100% correct but the expected has 1 error. Seems like an oversight. In the next bench, Sol looks like it’s correct again but the bboxes are rotated 90 degrees for some reason.
110.
by kzrdude 2mo ago
Nils van der Poel was* a long distance speed skater who set impressive records and wrote about his training regime. His book is https://www.howtoskate.se/ . He trained ambitiously for 3 years, won Olympic gold medals and ret
111.
by kzrdude 2mo ago
I think (maybe I missed something) that identical size and quant versions of Qwen 3.5 and 3.8 should run at the same speed. It’s the exact same architecture.
112.
by kzrdude 2mo ago
Some weeks ago a new official Gemma 4 release was posted that corrected some of the chat template problems. So the official release files on hugging face should be the way to go.
113.
by kzrdude 2mo ago
If they don’t loudly promise to not scan or store requests they handle then I believe they already do it, or will.
114.
by kzrdude 2mo ago
Depends on how you work with it. Reading the meandering and repetitive thinking is disturbing and taxing, so we can’t do that. So thinking has to be (mostly) hidden and just becomes waiting time.
115.
by kzrdude 2mo ago
It selects tokens but they expand to embedding vectors which are huge, also in memory and attention requirements, I think?
116.
by kzrdude 2mo ago
I don't get it, for pi.dev it's easy to replace or append to the system prompt using a text file. Per user or per project.
117.
by kzrdude 2mo ago
I would guess it's the 'vast' and the 'unprecedented' for example.
118.
by kzrdude 2mo ago
Is dflash still free there after the 0731 upgrade?
119.
by kzrdude 2mo ago
So what's needed to solve this is that someone publicly and popularly benchmarks using the chat template. So that they have an incentive to look good using it.
120.
by kzrdude 2mo ago
Is the chat template used at all when they benchmark the model?

More ›