Sam Pullara
Sam Pullara
Ready for testing. 1780ms on my machine :)
Updated and passes. BTW, having test.sh delete my 1B row file that I now need to regenerate is quite the troll :)
> Hum, hum, interesting point. I'd consider it volatile data, so 🤷 . The root issue is that we should have made the data set name an argument to the...
> Nice, looking good now. 00:12.063, super-clsoe #2 behind @filiphr with 00:12.027! His is using the hash code of the city as a key which is against the rules? Otherwise...
> Argh, again?! Seems I really need to add a test with 10K distinct keys. It's so easy to miss otherwise. Gnarf. You would have to construct a test with...
If they aren't comparing the actual string value of the keys you can construct keys that collide with their formula. You allow city names up to 100 bytes of UTF-8...
You should see how fast mine is if I use the same "optimization" :) The single largest chunk of time is spent comparing keys to ensure that they match.
I added changes similar to yours. Very good!
I took the maximum station name length of 100 as bytes (or some larger arbitrary number). If we choose that it will be easier to test and make sense of...
I'm loading CLIP image and text models and they are also extremely slow to load on ANE (CPU loads them instantly). 335M and 242M. On an iPhone 14 Pro Max...