Gracana

8 ms·

61.
by Gracana 7mo ago
Ha. I frequently visit this site to check out the latest little gadgets: https://liliputing.com/
62.
by Gracana 8mo ago
That's hilarious, I didn't realize you couldn't turn it off. I just tried disabling all the recommendation options and it still shows the category, except now instead of recommended items, it says "to show your recent fi
63.
by Gracana 8mo ago
Yeah, it’s the solvents used for cleaning the chambers and parts. Very nasty stuff, and it’s probably the biggest concern for this type of facility anywhere, not just in California.
64.
by Gracana 8mo ago
Awesome, thanks for the link.
65.
by Gracana 8mo ago
If the skill floor is high and therefore "effectiveness" is the same for a wide range of skill levels, isn't that the same as having a high barrier to entry? It seems that any activity or game where it takes a lot of skill be
66.
by Gracana 8mo ago
Huh, interesting. Do you know if the government sees the identity of the company and the person being verified? [edit] I did a little reading and it sounds like the company does not query the government with your ID. You get the cryptograph
67.
by Gracana 8mo ago
Where is it implemented that way?
68.
by Gracana 8mo ago
Are you sure about skill floor? I've only ever heard it used to describe the skill required to get into something, and skill ceiling describes the highest level of mastery. I've never heard your interpretation, and it doesn't
69.
by Gracana 8mo ago
They're talking about output consuming from the pool of tokens allowed by the subscription plan.
70.
by Gracana 8mo ago
If you enjoy that sort of thing, check out this guy's videos. Lots of trace repairs (including below the surface), pad replacements, etc. Quite impressive to see it done. https://www.youtube.com/@northwestrepair
71.
by Gracana 8mo ago
That's what I was expecting, too. Quite an interesting story regardless.
72.
by Gracana 8mo ago
> He made a social network for bots, FFS. Matt Schlicht made Moltbook, not Peter Steinberger.
73.
by Gracana 8mo ago
> I realize it makes you uncomfortable I think you're misunderstanding or misrepresenting them. The fight to have the most jaded or pessimistic take, the hottest flame, the spiciest rant, it's all so predictable and it's j
74.
by Gracana 8mo ago
There's a reddit comment here https://www.reddit.com/r/LocalLLaMA/comments/1r4m4it/comment... that says: my system is running GLM-5 MXFP4 at about 17 tok/s. That’s with a single RTX Pro 6000 o
75.
by Gracana 8mo ago
How? I think you’re imagining that these hypocrites exist.
76.
by Gracana 8mo ago
It's important to have enough VRAM to get the kv cache and shared trunk of the model on GPU, but beyond that it's really hard to make a dent in the pool of 100s of gigabytes of experts. I wish I had better numbers to compare with
77.
by Gracana 8mo ago
Yup, you're right.
78.
by Gracana 8mo ago
Six months ago I'd have said EPYC Turin. You could do a heck of a build with 12Ch DDR5-6400 and a GPU or two for the dense model parts. 20k would have been a huge budget for a homelab CPU/GPU inference rig at the time. Now 20k won
79.
by Gracana 8mo ago
Same. I tried really hard to quote properly, because I was so annoyed by the top-posting mess that everyone else did, and it frustrated me that people would add you to an email where you need to read 100 things that came before it (with inc
80.
by Gracana 8mo ago
Thanks for your review of five minutes of using Linux after 30 years of Windows use. Sorry to hear that zero effort didn't work out, but what can you do.. it's just not ready for the desktop. Enjoy Windows 11.
81.
by Gracana 8mo ago
A phone is a poor replacement for a laptop.
82.
by Gracana 9mo ago
Having dealt with the errata sheets for microcontrollers with all those fancy IO devices that solve real marketing problems etc, I'd kill to fix those problems with a software upgrade.
83.
by Gracana 9mo ago
You could have an LLM answer that, and then still interact as a human.
84.
by Gracana 9mo ago
That is a very good point and I would love to do it, but I built this machine in a desktop case and the motherboard has seven slots. I did a custom water cooling manifold just to make it work with all the cards. I'm trying to figure ou
85.
by Gracana 9mo ago
Same here. Disappointing. I wanted to run it on that picture of a church that looks like a chicken.
86.
by Gracana 9mo ago
I'm running the Q4_K_M quant on a xeon with 7x A4000s and I'm getting about 8 tok/s with small context (16k). I need to do more tuning, I think I can get more out of it, but it's never gonna be fast on this suboptimal ma
87.
by Gracana 9mo ago
What does scaling a person mean?
88.
by Gracana 9mo ago
I thought paging was so inefficient that it wasn't worth doing vs using CPU inference for the parts of the model that are in system memory. Maybe if you have a good GPU and a turtle of a CPU, but still somehow have the memory bandwidth
89.
by Gracana 9mo ago
I went looking for information on the oldest known lobster, and found this article about a 20lb lobster named George who was estimated to be 140 years old. Neat! https://en.wikipedia.org/wiki/George_(lobster)
90.
by Gracana 9mo ago
I'll have to remember that one, that's a good way to put it.

More ›