Toots for robbiemu@mastodon.social account

Written by Robbie on 2025-01-16 at 22:29

https://www.tiktok.com/t/ZP8FQuWEE/ What she said #tiktok #freespeech

=> More informations about this toot | View the thread

Written by Robbie on 2025-01-11 at 16:09

[#]FOSDEM Low-level AI Engineering & Hacking DevRoom is 2/2/2025 in Brussels. (I'm presenting in Quantization & research)

Here's what's on the agenda:

9:05am - 10:30 ML frameworks

10:30 - 11:20 Quantization & research

11:20 - 11:50 HPC-like compute needs & inference libraries

11:50am - 12:20am High-level use cases

12:20am - 1:40pm Orchestration & engine mgmt

1:40 - 2:40 LLM training & finetuning

2:40 - 4:20 Hardware: porting/optimizing ML frameworks

4:20 - 5:00 Inference on the edge

[#]FOSDEM2025

=> More informations about this toot | View the thread

Written by Robbie on 2025-01-11 at 14:53

[#]tiktok anyone know the api? I'd like to download my favorites, even as I don't care about my posted videos.

TikTok's lawyers argued that the ban will shut down the company, but I am hoping that will only be temporary and I can use it via VPN (you know, like any good Chinese citizen that uses google), even though I'll probably also need to create a new account not tied to my citizenship.

=> More informations about this toot | View the thread

Written by Robbie on 2025-01-01 at 13:45

[#]adventofcode #rust finally wrapped up part 1 with z3

I did not get 50 stars this year because of one day, day 21. I wanted to do it with a solver - which is a bad idea really. After ~1.2k line of code and 2k lines of tests, i have finally completed pt1 :) https://github.com/robbiemu/advent_of_code_2024/tree/day-21/day-21/src i haven’t merged it into main yet, and will need to take a different path in the end with part 2. It was a lot of fun, i even made a small helper lib to make some binary operations more ergonomic.

Happy holidays!

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-26 at 04:32

Day 25 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/25

my day25 #rust https://github.com/robbiemu/advent_of_code_2024/tree/main/day-25/src

Merry Christmas everybody :)

my progress this year: only day 21 kept me out this year.. I had decided that I preferred to solve it with a closed form solver and.. it's complex to do it this way. I'll probably come back in some days with those solutions

=> View attached media

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-26 at 03:18

I've completed "Crossed Wires" - Day 24 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/24

mine #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-24/src/lib.rs

this one would probably have taken me a lot of time normally (like day 21), but I was frustrated and peaked around a bit and noticed people were debugging the registers :D

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-25 at 20:36

I've completed "LAN Party" - Day 23 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/23

mine in #rust is just a plain petgraph solution https://github.com/robbiemu/advent_of_code_2024/tree/main/day-23/src

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-25 at 19:10

I just completed "Monkey Market" - Day 22 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/22

merry Christmas everyone! have some down time today so wrapping up ones I missed at the end.

mine is in #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-22/src/lib.rs

Kind of the opposite of performant: 8ms/430ms

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-21 at 13:55

I've completed "Race Condition" - Day 20 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/20

mine in #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-20/src/lib.rs

For the second day in a row, I had a solution that would generate actual results when we are expected to simply count them. Not quite bug free sadly, that is what prompted me to change approaches, but the debug printing the actual paths is pretty cool so I saved it https://github.com/robbiemu/advent_of_code_2024/tree/day-20-explicit/day-20

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-19 at 20:16

wondering if any #rust heads know why my benches vary so much wit this 1?

cargo bench --bench bench

...

bench fastest │ slowest │ median │ mean │ samples │ iters

╰─ main_bench 230.2 µs │ 41.42 ms │ 295.6 µs │ 713.1 µs │ 100 │ 100

cargo bench --bench bench --features part2

...

╰─ main_bench 225.7 µs │ 37.22 ms │ 299.6 µs │ 673.3 µs │ 100 │ 100

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-19 at 19:45

I just completed "Linen Layout" - Day 19 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/19

my #rust confusion, er, solution. https://github.com/robbiemu/advent_of_code_2024/blob/main/day-19/src/lib.rs

Started out avoiding recursion.. which is slow.

Got both down to ~310μs with some refactoring once I realized the counting mistake.

just want to point out, because I spent a couple of hours trying to actually generate a collection of each solution to each in part2, and you should not do that! Just count them!

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-19 at 01:21

it was bugging me that part2 was 340ms when part1 was 400μs. I realized that binary search was better than just testing all of them until. now I'm happy:

cargo bench --bench bench --features part2

Compiling day-18 v0.1.0 ...

bench fastest │ slowest │ median │ mean │ samples │ iters

╰─ main_bench 511.4 µs │ 793.2 µs │ 524.3 µs │ 540.6 µs │ 100 │ 100

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-18 at 21:45

This one I took a long time trying out rusty_planner's dstar_lite and later caboose crates, just because of how I imagined the problem should have been solved. It was fun and I learned some things.. I might submit a PR to rusty_planner, just to accept closures instead of functions.

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-18 at 21:42

I just completed "RAM Run" - Day 18 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/18

mine in #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-18/src/lib.rs

It is hard to argue with reusing nearly sufficient code from just a couple days ago :) (but that made part2 ~340ms)

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-18 at 03:03

I just completed "Chronospatial Computer" - Day 17 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/17

I had no patience for pt2 :D but by evening, it is done

mine #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-17/src/lib.rs

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-16 at 03:04

I just completed "Warehouse Woes" - Day 15 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/15

mine #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-15/src/lib.rs

relatively lightweight part 1, part2 got into the ms though:

╰─ main_bench 323.7 µs │ 568.7 µs │ 349.7 µs │ 356.4 µs │ 100 │ 100

╰─ main_bench 2.001 ms │ 2.659 ms │ 2.029 ms │ 2.076 ms │ 100 │ 100

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-13 at 23:59

I just completed "Claw Contraption" - Day 13 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/13

Thank god for an easy one after that monster part2 yesterday :D

my #rust solution https://github.com/robbiemu/advent_of_code_2024/blob/main/day-13/src/lib.rs

Far and away my slowest solutions yet, but I'm just orchestrating the z3 solver :D NOT MY FAULT haha

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-11 at 15:30

I just completed "Plutonian Pebbles" - Day 11 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/11

finally an easy day again! (mine: #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-11/src/lib.rs) This is again naive, didnt even use a cache til part2 -- for part 1 I actually ran the automata fully each step and returned the result! ~15ms for each part in benches

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-10 at 22:45

I just completed "Hoof It" - Day 10 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/10

(mine: #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-10/src/lib.rs ) this is basically lazy multi-source bfs -- I think recursion is the way to go, my runtimes are ~900μs

=> More informations about this toot | View the thread

Written by Robbie on 2024-12-10 at 00:41

I just completed "Disk Fragmenter" - Day 9 - Advent of Code 2024 #AdventOfCode https://adventofcode.com/2024/day/9

(mine #rust https://github.com/robbiemu/advent_of_code_2024/blob/main/day-9/src/lib.rs

I got a very performant O(n) solution for part1. I tried a couple of ways around getting part2 to also be O(n), but I key finding subtle issues that meant random access in a hash or else sorting, ie, O(n log(n)), so I just rewrote it to sort. 45μs for part 1, 25ms for part 2 -- probably both could have been faster, but for me, this was made me proud.

=> More informations about this toot | View the thread

=> This profile with reblog | Go to robbiemu@mastodon.social account