rolisz

27 ms·

391.
by rolisz 4y ago
I've been running into egregores (or at least the concept of egregores) a lot lately!
392.
by rolisz 4y ago
Fun party trick: look up old photos of people when you meet them after a long time and show it to them. Everyone is shocked/embarrassed. Extra points if you can pull up childhood photos of them.
393.
by rolisz 4y ago
But then who is Signal for? Yes, it has E2EE encryption, but who cares about that other than us orangey types? And Signal makes plenty of user hostile moves: making backups is not supported. If your Android phone dies, you can kiss your pre
394.
by rolisz 4y ago
No, but some people process different food differently, so eating the same amount of food might lead to different weight in different people. And no, the jury is not out on what food is healthy. Carbs are public enemy number 1 according to
395.
by rolisz 4y ago
Hey, I used to make that joke all the time (even when I was normal weight)!
396.
by rolisz 4y ago
Sure, but SOTA on NER is around 90% for things like Names and Places. For skills, a Spacy model will get around 60-70%. And training a Spacy model takes a bit of fiddling.
397.
by rolisz 4y ago
Nice! Are you going to release it publicly?
398.
by rolisz 4y ago
On a 1080Ti (so a 6 year old GPU), the large model runs in 1x time (so transcribing 10 minutes takes 10 minutes) and I've successfully transcribed even 1h+ files.
399.
I Will Fucking Dropkick You If You Use That Spreadsheet(ludic.mataroa.blog)
35 points by rolisz 4y ago|24 comments
400.
by rolisz 4y ago
I left Google in 2018. I was working in Zurich and moved back to Romania. My compensation dropped by ~60%. I don't regret the move at all. There's a lot of cool stuff outside of FAANG and there's lots of cool stuff outside of
401.
A case for Rust in deep learning(burn-rs.github.io)
1 points by rolisz 4y ago|0 comments
402.
by rolisz 4y ago
But what test do I write? What does the poorly documented 3rd party API return? Is the test failing because there's an edge case in the API or because my code is wrong?
403.
by rolisz 4y ago
I think the argument is that static methods are just functions, they are not strongly tied to a class or it's objects, so then let's not have two kinds of entities (normal functions and static methods) that do the same thing. Clas
404.
Making a loudness monitor for online meetings(rolisz.ro)
70 points by rolisz 4y ago|42 comments
405.
by rolisz 4y ago
I got many complaints from my wife that I'm speaking too loudly in online meetings, so I built a script in Python to warn me when I do speak too loudly.
406.
Europe Is Bracing for a Sharp, Abrupt Real Estate Reversal(bloomberg.com)
6 points by rolisz 4y ago|0 comments
407.
by rolisz 4y ago
Eh, in Romania mosquitoes are pretty bad. And in the last couple of years they're "cousins" came along: tiger mosquitoes. Those cause a really nasty bite
408.
by rolisz 4y ago
I've tried V60s, high-end Espressos, still hate the bitter taste of black coffee. Any other suggestions?
409.
by rolisz 4y ago
Sure, black coffee is fine, but I hate the taste of coffee, so I drink it with milk and Monin syrup
410.
by rolisz 4y ago
What accounting service do you use?
411.
by rolisz 4y ago
Because having something priced as "free" is distorts the market. Consumers don't see immediately the price they pay for the product. There's plenty of experiments with people having to choose between something that is $
412.
by rolisz 4y ago
Differential equations took a couple of years to grok. I first encountered them in high school, while preparing for physics Olympiad. I could solve basic differential equations by following the "rules" (such as for dampened oscila
413.
by rolisz 4y ago
Improve my nutrition. I want to lose weight, but I first need to fix my relationship with food. To do this, I plan to cook food myself and reduce preprocessed food.
414.
by rolisz 4y ago
Plenty of people live on farms and grow animals, kill them and then eat them. I haven't killed one myself, but I have been at pig butchering "parties" at my grandparents or uncles and saw how the sausage is made. Still tastes
415.
by rolisz 4y ago
Do you find it impressive when people get Doom running on a toaster? Or Doom inside Doom? This impressive on that level.
416.
by rolisz 4y ago
I think that is one of the benefits of having an easily parallelizable model.
417.
by rolisz 4y ago
Then you introduce a resizing component, which resizes differently for different parts of the image. And you still have to decide how to select the tile sizes: if it's heuristics, well, the field of ML has been moving away from that. I
418.
by rolisz 4y ago
I think that is not done because it's hard/unclear how to handle patches that are not the same size. In text it doesn't matter because all tokens are mapped to an integer (eg. the -> 54, hacker-> 1356).
419.
by rolisz 4y ago
Byte pair encoding (used in GPT2 models) is not semantic, it's purely based on statistics of occurrences of characters groups. It's not supposed to be semantic, it's just supposed to enable a complete coverage of the whole te
420.
by rolisz 4y ago
What about all the other countries in the world, where poor still exist, even if there is a public health care system?

More ›