nisten

9 ms·

61.
by nisten 2y ago
I am surprised at how much this thing is just straight up crushing it with just 8 cores. I think it topping the machine learning benchmarks has to do with having only 8 cores to share the 96MB of L3 cache, which ends up having a ratio of 1c
62.
by nisten 2y ago
Hmmm you might be able to do both no? Like the dedube is gonna run at the filesystem level but your memory security & ownership stuff is gonna run more efficiently. I am not sure.
63.
by nisten 2y ago
can someone smarter than me explain what happens when instead of the regular 4kb block size in kernel builds we use 16kb or 64kb block size or is that only for the memory part, i am confused. Will a larger block size make this thing good or
64.
by nisten 2y ago
This sounds about right from my experience getting nerdsniped by new samplers along with trying to reproduce the API middleware for the whole reflection thing, and using 4400 questions for a new benchmark is not bad given that even the well
65.
by nisten 2y ago
It's pretty interesting that the new SpinQuant method did not manage to be better than good old nf4bit QLORA training (Tim Dettmers really cooked with that one). Really appreciate that Meta published both results+model quants and didn&
66.
by nisten 2y ago
For a company selling intelligence, that's a pretty stupid way of labelling a new product.
67.
by nisten 2y ago
Is hackernews..hacked?
68.
by nisten 2y ago
Has anyone here actually ran the code on their own hardware yet? I did a standard non-middleware lm_eval_harness and got 0.3214 on gpqa_main_zeroshot WITH the systemprompt and 0.3616 without the systemprompt. Have not ran it with the middle
69.
by nisten 2y ago
it was a free opensource model release, the api was not for sale, there are literally over a million FREE models on huggingface.
70.
by nisten 2y ago
There is no official API, you're confirming a temporary one that was taken down weeks ago. That was done via OpenRouter, a routing api service/site, which routes to different model's under load. Yes they could've switche
71.
by nisten 2y ago
yes and the massive increase to GPQA scores from o1 was attributed to this technique, so there is something there, despite the hard feelings of unproductive reddit users
72.
by nisten 2y ago
they're 140Gb folders, for each checkpoint, yes file corruption happens and as for the fraud part...it was an opensource model release that did not meet the claimed benchmarks when people tried to replicate it
73.
by nisten 2y ago
Because it's a law that first intended to put opensource developers in jail.
74.
by nisten 2y ago
Imagine being concerned about AI safety and then introducing a bill that had to be ammended to change criminal responsability of AI developers to civil legal responsability for people who are trying to investigate and work openly on models.
75.
by nisten 2y ago
looking at the package.json infestation of this thing makes me want to just go plain vim WITHOUT any plugins. Ok, what's that , you want AI sir, guess what b:tch, you can :split out a :terminal window and run a llama-cli conversation r
76.
by nisten 2y ago
they got old
77.
by nisten 2y ago
Machine learning is extremely good at recognising patterns and I'd much rather trust an LLM's spotting accuracy for an early warning system than the regex code of hospital IT workers
78.
by nisten 2y ago
No there isn't, this is a 4 year old project that likely started before back-n-forth prompts were really a thing.
79.
by nisten 2y ago
How much RAM?
80.
by nisten 2y ago
These types of takes seem pretty ignorant of the percentage of workers and students that are using artificial intelligence tools daily. It is ignorant because it ignores any economical improvement that comes from say a machinist figuring ou
81.
by nisten 2y ago
Complete horseshit marketing speak. Was the cloud non-private before? Was it not secure in the first place? Do my Siri searches no longer end up as google ads metadata now? Are the feds no longer able to get rubber stamp access to my i C L
82.
by nisten 2y ago
I thin it's the other way around, most western companies, especially older small businesses are actually high-trust environments and treat employees like family, or at least try to. I'd make the argument that authoritarian civiliz
83.
by nisten 2y ago
Getting the wrong vibes from the dark patterns on this site. I click on github.. it doesn't open github... had to look in the comments here. I find repo here... the front end is some ancient angular app with 3 year old security vulnera
84.
by nisten 3y ago
Meh, the comparison is somewhat pointless when it doesn't account for the slowdown that the vast majority of pytorch codebases experience on TPU's vs using JAX and it's accelerations specific to TPUs, and vice versa. https:
85.
by nisten 4y ago
Long live YouTube. Long enough until some client-side stable-diffussion thingy for generating tutorial videos becomes practical enough for daily use.
86.
by nisten 4y ago
It looks like dexie supports svelte, this is pretty awesome thank you. https://dexie.org/docs/Tutorial/Svelte
87.
by nisten 4y ago
I had not looked into RxDB before thank you. By any chance you know how much their Premium plugin with IndexedDB support costs, their price is not listed.
88.
by nisten 4y ago
I think it really was the M1 Chip that killed Mighty. The product itself looked like it was working great. Before the value proposition was good. You can still use your old laptop and still have a very fast browser experience via remotely
89.
by nisten 4y ago
Logically it's not a deal breaker (as Electron is quite mature) but practically there are a few problems: Users are downloading apps less and less and relying more on web links from apps they already have. So they mostly tend to end up
90.
by nisten 4y ago
It's rare to see offline support in a web app for a few reasons: LocalStorage is easy to use but too limited at 50Mb. WebSQL got deprecated by FireFox because of secuirty issues years ago ( albeit is still somewhat supported in Chrome

More ›