3,132 karma · joined March 15, 2017
MoE models are kind of bonkers.
Physical violence is definitely going to leave lasting effects, but words can still do the same.
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaẞ
Boy I sure do wonder what character could possibly be causing issues there. whereas without shrinking it might instead spit out something like
ЁЂЃЄЅІЇЈЉЊЋЌЍЎЏАБВГДẞЕЖЗИЙКЛМНОПРСТУФХЦЧШЩЪЫЬЭЮЯабвгдежзийклмноп
Which makes finding the offending character a lot harder.
If you get hit by a tram it is because you are in a place where you should have known the tram would hit you. The things are on tracks.
needle_position = needle_position * 0.9 + desired_needle_position * 0.1
is very easy to implement, easier than linear if you ask me, and should give it a nice damped movement. The constants can be tuned to go as quick or as slow as you like of course.
:= was bolted on after of course, and I think that was a mistake, but no one asked me, nor do I expect anyone to ;P
In general the formulation is more important than the brand, and the formulation isn't /that/ important either. If it's lead free, has a rosin core, and comes on a spool, you can probably use it.
if is_logged_in=True:
allow_access()
I've got... opinions on the way python separates statements and expressions, but there's some real benefits to it too.it really isn't if you use a nice modern lead-free solder. you'll need your iron to be about 20c hotter, but it's not like the early days of lead-free where it'd flow all weird.
Still wash your hands after using lead-free solder by the way. You don't want to be eating rosin or copper either.
---
Waldo is located in the *center* of the image.
Here is how to spot him:
1. Look towards the middle of the beach scene.
2. Find the *red and white striped lifeguard stand/booth* on the right side of the center area.
3. Just to the left of that booth, there is a man standing up wearing his signature *red and white striped shirt*, blue jeans, and red hat with a pom-pom. He also has a camera around his neck.
---
Which is not generally something OCR can do. If you instead ask "how many horses are in this image?" you get this response:
---
Based on a careful look at the image, there are *3* equine animals (which appear to be donkeys or ponies used for beach rides).
1. One is near the top left, being ridden by someone in yellow shorts.
2. Right next to it is another one, being ridden by someone in blue and yellow.
3. On the far right edge of the image (near the water), there is a third grey animal being ridden by someone wearing a hat.
--
Now is this all anything you can't do with more boring machine learning? sure, but there's something incredibly convenient about how generic LLMs are. You don't need to train anything, just point the LLM at an image and ask.
[0] https://i.pinimg.com/originals/18/64/44/1864444c819a7adae742...
D's equivalent to "constexpr" is "compile time function evaluation". i.e. in any context where it only makes sense to run code at compile time, it will do so. This makes it trivial to do some pretty complex things at compile time. I put together an example that shows creating static arrays, dynamic arrays, and a dynamic array with a partial fibbonaci sequence all at compile time[0].
[0] https://gist.github.com/SuperDoxin/d9fcc68b73c035cbde7f0bd08...
yeah if you're aiming to make a proper professional game with aims of making profit... the playdate probably isn't the way to go. But then I recon that's part of what makes it an awesome platform. It hasn't been captured by capitalism yet.