_0ffh

14 ms·

121.
by _0ffh 2y ago
As a rule of thumb, the bigger the model is, the more graciously it degrades under quantisation. So you may assume performance loss for a 8B model would be lower than for a 3B model. (I know that doesn't make up for missing numbers in
122.
by _0ffh 2y ago
Thanks yes. Though I did not mean the bare possibility, but intended use case, which may lead to different design choices.
123.
by _0ffh 2y ago
I guess he might mean this one https://shader-slang.org/ though at first glance at least it looks more graphics than GPGPU oriented. Edit: Hmm, this part of the same project looks general purpose-y and apparently integrates
124.
by _0ffh 2y ago
Similar here. What the teachers where actually looking at was if the calculations and error analysis were done right.
125.
by _0ffh 2y ago
The couple who runs the website is half native Chinese and have lived and traveled there a lot. Source: Been watching their YT channel for a while.
126.
by _0ffh 2y ago
Also, the guide itself is accompanying material to this cooking channel: https://www.youtube.com/@ChineseCookingDemystified Interestingly, they have referenced chef Wang a couple of times, among few other YT channels.
127.
by _0ffh 2y ago
Who else clicked this, thinking it might be about music?
128.
by _0ffh 2y ago
Another thing why magnesium oxide could not be the optimal compound for you: It's a laxative. That might be good to know before you start daily supplementation. Personally, I have bisglycinate.
129.
by _0ffh 2y ago
Please don't confuse people with wrong information, the reasoning part in reasoning models is the exact same LLM that produces the final answer. For example o1 uses special "thinking" tokens to demarcate between reasoning and
130.
by _0ffh 2y ago
Not a valid one at least.
131.
by _0ffh 2y ago
> This enormous acceleration of progress is increasingly unnoticed due to even faster increases in software bloat Can't resist to throw in this quote (from one of Wirth's students I think): "Software gets slower faster tha
132.
by _0ffh 2y ago
I am sorry, I don't quite understand where the "abuse" comes in. Attaching function pointers to structs? I've done that in production twenty years ago (and others have probably done it in the 70s) and I see nothing wrong
133.
by _0ffh 2y ago
Okay, the part of "free markets are possible without capitalism" is quite obvious. But going back to the earlier messages, what happened is that I assumed that capitalism implies free markets (not the other way around), and that&#
134.
by _0ffh 2y ago
> i think you read way too much into what i said Well, it seems like I really may have done that. I actually hadn't even noticed that my interlocutor had switched inbetween. > i'm just saying capitalism != free trade Right,
135.
by _0ffh 2y ago
The hammer allows the worker to be more productive than he would be without one, so he chooses freely to team up with the hammer-owner for a share of the additional value created by using the hammer. The endlessly repeated "but all muh
136.
by _0ffh 2y ago
Do you own a hammer? Well, that's a capital good, so you're a capitalist! Congratulations!
137.
by _0ffh 2y ago
> but doesn't, because he's a nice person He doesn't, because he would quickly loose all his business to the competition, and if there was none, there quickly would be - unless government force provides him with an artific
138.
by _0ffh 2y ago
Capitalism posits that it should be legal to privately own capital goods. If I own something, that makes me the one who decides what to do with it.
139.
by _0ffh 2y ago
> I don't care about that level of penny-ante pedantry. That's turns every forum into hyper-correctionalist tedium. And I don't care for sloppy thinking. It leads to all kinds of bad conclusions. > As I already quoted,
140.
by _0ffh 2y ago
> the large majority of laissez-faire capitalists still want a government That may be true, but even a majority doesn't make it true that "Laissez faire capitalists still want a government". You'd have to prepend a
141.
by _0ffh 2y ago
> Laissez faire capitalists still want a government No, not as a rule, they don't. Some may, but even those must be against government interference in the market, as that's the definition of laissez faire. The only relevant div
142.
by _0ffh 2y ago
Without government force to back it up, who would care?
143.
by _0ffh 2y ago
> I don't know, I view capitalism mostly as a way to allow people who have money to make more money without providing work or value from people who do work Well, that's not how capitalists view it, quite the opposite. In a free
144.
by _0ffh 2y ago
> The solution is file sharing (via BitTorrent & co), which is very much not capitalism. You'd be surprised, how many laissez faire capitalists regard "intellectual property" to be an anti-capitalist artificiality. The
145.
by _0ffh 2y ago
Maybe I'm not so bad for using "superfluous" braces quite often, although this is not specifically the reason why I do.
146.
by _0ffh 2y ago
Why go to techcrunch and not directly to their only source of information on this? There are also some actual technical details there. https://www.nvidia.com/en-us/ai/cosmos/
147.
by _0ffh 2y ago
The context first bit doesn't even make sense. Transformers process their whole context window in parallel, unlike people who process it serially. There simply is no place that gets processed "first". I'd love to see a
148.
by _0ffh 2y ago
Very sceptical of "Context First: We front-load system understanding before diving into code". The LLM sees the whole input at once, it's a transformer, not a recurrent model. Order shouldn't matter in that sense. Ed. I
149.
by _0ffh 2y ago
Something in this vein has actually been made quite some time ago: https://www.bltt.org/software/dasher/
150.
by _0ffh 2y ago
Link doesn't work for for me (not found error), but this one does (including pdf d/l): https://www.researchgate.net/publication/385937685_Journal_o...

More ›