ipsum2

46 ms·

691.
by ipsum2 3y ago
probably mean agnostic, agonistic implies the opposite.
692.
by ipsum2 3y ago
Doesn't work on image classifiers, because there's no KV cache. Also, standard image classifiers can do 100-1000 images/sec without any optimizations.
693.
by ipsum2 3y ago
The ideas are orthogonal, and can be used (theoretically) at the same time.
694.
by ipsum2 3y ago
What happens if someone searches their home address or a place nearby? If it's automatically cached, it could be a data leak. Some sufficiently motivated person can correlate it with someone who connects to Mullvad servers.
695.
by ipsum2 3y ago
The issue might be more interesting to read: https://github.com/easylist/easylist/issues/16372
696.
by ipsum2 3y ago
That sounds highly implausible. Bard would not be trained on internal Google code, nor would have access to it. Most likely it was hallucinating, based off of public APIs its seen. Did you ever get confirmation that that code actually exist
697.
by ipsum2 3y ago
This shouldn't happen if you use pip to install packages, not conda.
698.
by ipsum2 3y ago
Use conda to set up new environments and switch between them, and use pip to install modules. It's really not as complicated as people make it seem. I haven't had any issues with this set up across Linux and MacOS, doing web &#x2F
699.
by ipsum2 3y ago
I suspect GPT-4 has some magic that's not only the original RLHF using PPO. Either an augmented dataset or some enhanced version of RLHF.
700.
by ipsum2 3y ago
You can also make this with a box fan, construct a box with 4 filters of roughly the same size as the fan, the bottom of the box can be cardboard. Did this back in 2019 and it works great, though I ended up switching to a Coway a few years
701.
by ipsum2 3y ago
Maybe, it depends on the lens design as well.
702.
by ipsum2 3y ago
It's a research paper, not a product. I don't know why you're against this, but I appreciate researchers that publish papers so knowledge isn't siloed in organizations and contribute to the public good and advancement of
703.
by ipsum2 3y ago
Invalid comparison, the display size doesn't matter because of how lenses work, so the perceived screen size is what counts. VR headsets are typically measured in PPD (pixels per degree). Apple's headset is rumored to be 1.5-2x Qu
704.
by ipsum2 3y ago
The user kyegomez writes shitty, incorrect implementations of popular papers to obtain clout.
705.
by ipsum2 3y ago
That's for inference, I'm referring to training.
706.
by ipsum2 3y ago
It's because ROCm is not developed for RDNA (consumer) cards, but CDNA (datacenter) cards. No surprise that he's having trouble with it.
707.
by ipsum2 3y ago
I have some experience in this area, having both worked on machine learning frameworks, trained large models on datacenters, and have my own personal machine for tinkering around with. This makes very little sense. Even if he was able to ac
708.
by ipsum2 3y ago
Meta has had a culture of open source, even before their reputation got bad (React, RocksDB, Thrift, folly, etc). They open source a lot of their AI models (instead of just source) because they don't have an enterprise cloud to sell th
709.
by ipsum2 3y ago
There's no explanation about why the OP thinks the way he does. If there was some interesting commentary or question, that would be worth reading about.
710.
by ipsum2 3y ago
What's the point of posting this? You may not be their target demographic. No one's forcing you to use it. If I see a post about a new Lisp dialect (something I have no interest in), commenting "That will be a hard no from me
711.
by ipsum2 3y ago
The people who care about decentralization and like Meta/Facebook are two separate circles, but its interesting that they're supposedly building on ActivityPub.
712.
by ipsum2 3y ago
Webtoons (a Korean company) is probably more popular than manga now for this reason.
713.
by ipsum2 3y ago
You're looking for Koomey's Law, which is a different concept than Moore's law.
714.
by ipsum2 3y ago
Why do you feel the need to appeal to authority here? Moore's law states that the number of transistors on a chip doubles every two years. Looking at the data[0] confirms, yes, Moore's law is still alive. [0]: https://u
715.
by ipsum2 3y ago
I don't think its a template, it appears to be some LLM. Not sure if its chatGPT or something else.
716.
by ipsum2 3y ago
Those are made-up numbers by Nvidia. Obtaining anywhere close to that in reality is basically impossible. Better to compare benchmarks instead.
717.
by ipsum2 3y ago
GPT-4 instruction dataset
718.
by ipsum2 3y ago
xpos is SoTA right now: https://arxiv.org/pdf/2212.10554.pdf
719.
by ipsum2 3y ago
I've been playing with StarCoder for the last week. It performs great once fine-tuned. Highly recommend people use it as a base model for anything, not just coding.
720.
by ipsum2 3y ago
It's not clear where to draw the line on what is AI or not, and I suspect most content will be at least partially generated by AI. Is spell-checking AI? grammar-checking? Rewriting sentences for clarity? Summarization?

More ›