Tootfinder

Knocked ✊
敲 ✊
📷 Nikon FE
🎞️ Ilford FP4 Plus 125, expired 1993
If you like my work, buy me a coffee from PayPal #filmphotography

I mean, it's not a bike so am I really breaking any rules? 🤷
#unicycleLife #Oslo #Operahuset

EU-Vorschlag: Google soll Such- und Gemini-Daten rausrücken
Die Europäische Kommission hat Google Vorschläge übersandt, wie das Suchangebot wettbewerbsrechtskompatibler werden soll.
https://www.

OK, I know I keep bragging about #Clojure being remarkably free of bit-rot, and it is.
But I've just hit bit-rot in an eleven year old project of mine, and I'm not sure what to do.
The error is in a #leiningen project built on a

So to follow up on this, I've caught it in action. Models, when quantized a bit, just do a bit more poorly with short contexts. Even going from f32 (as trained) to bf16 (as usually run) to q8 tends to do okay for "normal" context windows. And q4 you start feeling like "this model is a little stupid and gets stuck sometimes” (it is! It's just that it's still mostly careening about in the space of "plausible" most of the time. Not good guesswork, but still in the zone). With long contexts, the probability of parameters collapsing to zero are higher, so the more context the more likelihood you are to see brokenness.
And then at Q2 (2 bits per parameter) or Q1, the model falls apart completely. Parameters collapse to zero easily. You start seeing "all work and no play makes jack a dull boy” sorts of behavior, with intense and unscrutinized repetition, followed by a hard stop when it just stops working.
And quantization is a parameter that a model vendor can turn relatively easily. (they have to regenerate the model from the base with more quantization, but it's a data transformation on the order of running a terabyte through a straightforward and fast process, not like training).
If you have 1000 customers and enough equipment to handle the requests of 700, going from bf16 to q8 is a no-brainer. Suddenly you can handle the load and have a little spare capacity. They get worse results, probably pay the same per token (or they're on a subscription that hides the cost anyway so you are even freer to make trade-offs. There's a reason that subscription products are kinda poorly described.)
It's also possible for them to vary this across a day: use models during quieter periods? Maybe you get an instance running a bf16 quantization. If you use it during a high use period? You get a Q4 model.
Or intelligent routing is possible. No idea if anyone is doing this, but if they monitor what you send a bit, and you generally shoot for an expensive model for simple requests? They could totally substitute a highly quantized version of the model to answer the question.
There are •so many tricks• that can be pulled here. Some of them very reasonable to make, some of them treading into outright misleading or fraudulent, and it's weirdly hard to draw the line between them.

Series C, Episode 13 - Terminal
ZEN: The continuing malfunction of hull sensors is limiting data for analysis. Operational systems indicate that the damage is widespread but superficial.
DAYNA: The nature of the damage. What's causing it?
https://blake.torpidity.net/m/313/255 B7B4

For a year, the Trump Justice Department has been on an odd mission:
to assist a former FBI informant with ties to Russian intelligence who ended up in prison for passing disinformation about Joe Biden to the bureau.
His crime deeply affected American politics.
The false claim he slipped to the FBI
—that Biden and his son Hunter each were paid a $5 million bribe by a Ukrainian energy company
—became the main evidence in the House Republicans’ reckless and ill-fate…

Trump’s DOJ is helping a convicted FBI informant tied to Russian intelligence
Alexander Smirnov’s lies about Joe Biden were at the core of the GOP’s impeachment drive.

One of the biggest current myths is that if democrats win coming midterms and presidential elections everything will be allright and we can return to business as usual with the USA. In my view that is not the case, the political system and society is too broken to be healed that quickly after Trump's destructive presidency. It will take many more more years maybe even a decade to heal and reform.
#Trump

Boah, bin ganz "schön" #Pollen-kontaminiert
Kann ich das #Prednisolon'20 mg noch net absetzen
https://

Was just looking for something on our local Bargains Hunt style auction shop, and noticed a link on the nav menu: "AI videos". Well, I had to look, didn't I... These are TOTALLY NOT FREAKY at all, there is something wrong with your head if you think otherwise... Brainworms, probably.
https://www.mewsauctions.co.uk/ai-vi…