search

krukah / robopoker

Play, learn, solve, and analyze No-Limit Texas Hold Em. Implementation follows from Monte Carlo counter-factual regret minimization over with hierarchical K-means imperfect recall abstractions.
MIT License
43 stars 5 forks source link

Print progress for layers compute #18

Closed reinismu closed 1 month ago

reinismu commented 1 month ago

image

krukah commented 1 month ago

big fan of this @reinismu -- much cleaner than our existing Progress struct and impl. specifically it's good to have the ProgressBar be able to shared across threads, i never got the whole Send and Sync and 'static stuff down for my own implementation for it to be of any use in our parallelized workloads. so it's nice to have this crate to not reinvent the wheel.

i'd like to get around to taking these changes into the crate. it actually pairs quite nicely with a new lazy functional ObservationIterator i whipped up yesterday, since we can use that to calculate (compile-time const) Street::n_observations or Street::n_isomorphisms depending on whether i get around to "isomorphizing" the Layer projections, which i probably will pretty soon. both ObservationIterator::combinations() and Street::n_observations will work consistently, as verified in the cards::observations::tests module