redox99

36 ms·

751.
by redox99 3y ago
12GB/s seems a bit slow. I'd expect the only bottleneck to be memory bandwidth. A dual channel DDR4 system memory bandwidth is ~40GB/s, and DDR5 ~80GB/s. Since this operation requires both a read and a write, you'd
752.
by redox99 3y ago
> PID's do more harm than good In what way?
753.
by redox99 3y ago
> shouldn't the legal department be involved way before the issue reaches the engineer? Possibly, but the legal department might not know what cloudflare is or the implications until the engineers explain it.
754.
by redox99 3y ago
Yes, who would also yell at you when customers can't access their cat pictures. Of course it could be an accident, and you didn't know. But if it was clearly malicious compliance, personally and I think I speak for many, I would
755.
by redox99 3y ago
Obviously your boss or boss's boss would care.
756.
by redox99 3y ago
You can mute people who slur, you can't "mute" cheaters.
757.
by redox99 3y ago
Really depends on the game. I'm sure in Rust most people would go to the "unmoderated" server.
758.
by redox99 3y ago
I don't know what you mean by toxicity, but "back in the day" as you mention, you'd get called an n-word 5 times a match. Nowadays that doesn't happen in most games because they ban you if you do so.
759.
by redox99 3y ago
I don't really care the same way I don't care about the existance of certain websites or subreddits that I disagree with. I rather deal with moderation by myself by choosing who to follow, than rely on Twitter moderators and their
760.
by redox99 3y ago
llama-chat is censored, not base llama
761.
by redox99 3y ago
Really depends on how much you weigh CO2 emissions, different air pollutants, noise pollution, etc. Environmentally friendly isn't a single dimension. Furthermore it's not like he's going to dispose his older car, it's g
762.
by redox99 3y ago
If I had to guess, they probably don't fully top up.
763.
by redox99 3y ago
Thunderfoot is more focused on being a contrarian than being accurate and unbiased. See sibling comments that explain why it being a ceramic isn't that relevant.
764.
by redox99 3y ago
Sorry can't provide any resources right now. If you search a bit I'm sure you'll find some latency comparisons between DDR and GDDR. But basically GPU memory (GDDR5/6/6X/etc) is optimized for bandwidth (because
765.
by redox99 3y ago
Because 175B parameters (350GB for the weights FP16, let's say a bit over 400GB for actual inference), fit very comfortably on 8xA100 (640GB VRAM total). And basically all servers will have 8xA100 ( maybe 4xA100). Nobody bothers with
766.
by redox99 3y ago
You can find content like this on Twitter if you follow the right people. In fact I read this article before it was even posted here because @karpathy tweeted about it.
767.
by redox99 3y ago
VRAM is not latency optimized. VRAM has worse latency than your CPU RAM. The reason why it's mounted closer is because of signal integrity because of higher frequencies, not because of latency.
768.
by redox99 3y ago
>Still, we are looking at GPT3 level of performance on one server rack. That says something when less than a year ago, such AI was literally magic and only run on a massive datacenter. I'm not sure what you mean by this. You've
769.
by redox99 3y ago
Potentially it could have an impact if it omits a high level description before writing the code, although obviously things like "Sure! Happy to help" do not help. In practice I haven't seen it make too much of a difference
770.
by redox99 3y ago
Use a better model. airoboros supports the PLAINFORMAT token "to avoid backticks, explanations, etc. and just print the code". https://huggingface.co/TheBloke/airoboros-l2-70B-GPT4-2.0-GG...
771.
by redox99 3y ago
> it's very good for HN headlines to make a $X.cpp but it's not the right tool for products. Funny how it's the opposite. ONNX in this case, outside of the HN headline and saying "we did it" is almost useless. LL
772.
by redox99 3y ago
GPT4 made a mistake on its first try, but after asking what the complexity of pop is, it figured out its mistake and fixed it. https://chat.openai.com/share/d527f65f-8a6d-4602-acab-4d80ed...
773.
by redox99 3y ago
Obviously not. For example using Chain of Thought will increase the model's performance[1] [1] https://arxiv.org/abs/2201.11903
774.
by redox99 3y ago
Those are mid size
775.
by redox99 3y ago
Anecdotally I think it's much more common for people in the US to do some work on their house by themselves than in the EU.
776.
by redox99 3y ago
Pretty sad. They don't care about regulation that would potentially have a profound impact for everybody, but they care about people voluntarily doing something for a silly crypto token?
777.
by redox99 3y ago
Obviously they wouldn't just slap it in a box and send it through FedEx. The authors would personally deliver it.
778.
by redox99 3y ago
Source?
779.
by redox99 3y ago
Very misleading title
780.
by redox99 3y ago
Thus the two options

More ›