#diary #программирование #ml #покер
Мой чёртов RL-агент, над которым я тут сижу, наконец-то пробился через первую стену, проверяющую его скиллы.
Iteration 1 Epoch 140: num_players=8
Timing: iteration_elapsed=29400.5s, epoch_elapsed=162.5s; iter model timesteps=7168000, full model timesteps=7168000
ETA to end of 20-epoch block: 0s (2026-02-17 05:02:37)
Eval num_players=2: RivalType5: 756.7, Total: 756.7 BB/100; rl_clusters:
maniac=0.85, aggro_fish=0.68, calling_station=0.60, loose_passive=0.52, lag=0.51, fit_or_fold=0.25, weak_tight=0.18, tag=0.13, reg=0.08, nit=0.00, over_folder=0.00BB/100 estimation exceeded 10 minutes; stopping early.
Eval num_players=4: RivalType1: 91.2, RivalType3: -2.0, RivalType1-1: 73.2, Total: 162.3 BB/100; rl_clusters:
fit_or_fold=0.87, loose_passive=0.77, aggro_fish=0.60, calling_station=0.58, weak_tight=0.58, maniac=0.50, nit=0.46, over_folder=0.33, lag=0.27, tag=0.07, reg=0.00BB/100 estimation exceeded 10 minutes; stopping early.
Eval num_players=6: RivalType2: 7.8, RivalType1: 54.1, RivalType5: -8.7, RivalType1-1: 40.1, RivalType4: 0.7, RivalType4-1: 0.8, RivalType8: 0.9, RivalType7: 1.2, RivalType5-1: 2.0, RivalType3: 1.1, RivalType1-2: 2.3, RivalType6: 3.2, Total: 105.5 BB/100; rl_clusters:
loose_passive=0.81, calling_station=0.76, aggro_fish=0.60, weak_tight=0.59, fit_or_fold=0.58, nit=0.52, over_folder=0.33, maniac=0.25, reg=0.12, tag=0.09, lag=0.00BB/100 estimation exceeded 10 minutes; stopping early.
Eval num_players=8: RivalType1: 31.4, RivalType5: 8.3, RivalType1-1: 28.4, RivalType2: 30.6, RivalType3: 1.5, RivalType4: 1.5, RivalType6: 0.1, RivalType4-1: -0.9, RivalType3-1: -0.9, RivalType5-1: 1.2, RivalType2-1: 2.5, RivalType7: 2.5, RivalType1-2: 0.5, RivalType8: 0.2, Total: 106.9 BB/100; rl_clusters:
loose_passive=0.71, calling_station=0.66, aggro_fish=0.60, fit_or_fold=0.58, weak_tight=0.49, nit=0.34, over_folder=0.33, maniac=0.25, reg=0.24, tag=0.22, lag=0.02Saved poker_model_466780b9_0001.zip/.json/.py
Iteration 1 complete. All modes > 10.0 BB/100
Вижу это так: