3abiton

12 ms·

241.
by 3abiton 1y ago
I am curious about also the varying quantization of kv cache. It seems quantizing values yield better results than doing so to keys
242.
by 3abiton 1y ago
That's the issue eith AI nowadays, ehenever I see a good writeup or a project, I wonder bow much of it is AI.
243.
by 3abiton 1y ago
I'm curious about that syllabus.
244.
by 3abiton 1y ago
Great write up, it would be interesting to see a lot of those covered features in comparison to other frameworks!
245.
by 3abiton 1y ago
On that note, I wonder if having LLM agents communicating with each others in a human language rather than latent space is a big limitation.
246.
by 3abiton 1y ago
This an amazing solution. I wondered about this often, looking at you `uv`, but in a lot of the cases I cave given that everyone else trust some code maintainers.
247.
by 3abiton 1y ago
> MCP seems like a more "in-between" step until the AI models get better. I imagine in 2 years, instead of using an MCP, we will point to the tool's documentation or OpenAPI, and the AI can ingest the whole context without
248.
by 3abiton 1y ago
And they say our scale up is siloed. Leave it to google to show' em.
249.
by 3abiton 1y ago
What's their selling point compared ro Qubes?
250.
by 3abiton 1y ago
> Unlike traditional Transformer designs, which suffer from quadratic memory and computation overload due to the nature of the self attention mechanism, our model avoids token to token attention entirely. I skimmed the paper, and unlike
251.
by 3abiton 1y ago
I had a similar problem recently: I got this cheap 5.1 surround system soundbar that support up to Dolby TrueHD via HDMI. But here is the catch, it only works with eArc enabled devices (new gen TV). If you plug your PC you need to use SPIDF
252.
by 3abiton 1y ago
> A more concerning limitation is that when the city re-ran parts of its analysis, it did not fully replicate its own data and results. For example, the city was unable to replicate its train and test split. Furthermore, the data related
253.
by 3abiton 1y ago
To hammer one point though, you have to understand that researcher are desensitized to minor novel improvement that translate to great value products. While obviously studying and assessing the limitations of AI is crucial, to the general p
254.
by 3abiton 1y ago
It seems to be the case, although pytorch rocm is coming around slowly. Very slowly, if you get it working that is.
255.
by 3abiton 1y ago
When I worked in Demand prediction (multivariate), it was lgbm that was outperformong across the board.
256.
by 3abiton 1y ago
Interestingly the author mentions LoRa as a "special" way for fine-tuning thatis not destructive. Have you considered it or you opted for more direct fine-tuning?
257.
by 3abiton 1y ago
During my research years, we had to grind on Combinatorial Optimization book by Korte and Bygen for the weekly book reviews. Safe to say, it was not an introductory work. Still it was fun seeing the different examples my colleague would bri
258.
by 3abiton 1y ago
Thanks! I have already setup iptables rules for vms to deny them local network access. I'll use the same trick for local access now i guess.
259.
by 3abiton 1y ago
I still have the Claude subscription, so I will take a look again and see.
260.
by 3abiton 1y ago
I don't really measure it as I tune by taste these days. I usually soak a piece of lime/lemon overnight, and sometimes I add ginger or cumumber and a generous sprinkle of salt to a big water jar. I tried some of the other minerals
261.
by 3abiton 1y ago
I majored in CS and I had no idea that was possible: public websites you access have access to your local network. I have to take time to process this. Beside what is suggested in the post, are there any ways to limit this abusive access?
262.
by 3abiton 1y ago
> I found all the previous Gemini models somewhat inferior even compared to Claude 3.7 Sonnet (and much worse than 4) as my coding assistants. What are your usecases? Really not my experience, Claude disappoints in Data Science and compl
263.
by 3abiton 1y ago
Do you have any good practical guide in mind?
264.
by 3abiton 1y ago
Does that fly in the EU?
265.
by 3abiton 1y ago
> Anything that kicks off your metabolism will make it not fasting. That's the key part. Black coffee is key. I also sometimes go with tea or fruit infused water.
266.
by 3abiton 1y ago
It's like that saying from Nicholas Taleb(I onow he's controversial), to read news from a year ago, so realize how much bullshit is in news. https://medium.com/incerto/the-facts-are-true-the-news-is-fa...
267.
by 3abiton 1y ago
It's easier to succeed if you ignore the issues, andthe users are not aware of it.the rate of evolution of "AI" recently is so fast, no one is stopping to do actual benchmarks and analysis of allyhe new models.
268.
by 3abiton 1y ago
A reminder that it's possible to use tools like XPL-EX to circumvent those attempts. Also ad blocking via adaway would do the trick here I assume, as it should block Meta Pixel tracking. Overall, awful approach.
269.
by 3abiton 1y ago
Except it's quite tough to fight hidden wars in big teams. I was part of that hidden wars, where a team member would totally misinterpret a situation, goes full on complain to the manager, the latter lacking oversight over all of the t
270.
by 3abiton 1y ago
It depends how you do it. My current approach relies on intermittent fasting for 18 hours (save it for coffee and salty water), and accasional 24-36 hours fast (approx every 6-8 weeks). Not a keto diet, but helps with training Ketons.

More ›