liquidcat

I trained models though evolution training, all is left is to just make it useable with good inteface

yeah, so the issue is that LSTM’s are not fit to train through evolutionary technics, I remade encoder to return winrate and confidence of other players and implmeneted simple MLP which can be trained with this technic

I cahnged architecture of my model to have 2 heads, I also worked on fully working model, right now it just needs improvments in encoder part.

I changed env for one suited for table with more than 2 players, right now I am designin model and encoder for data, simulating poker is done and system of reward is done, only part left in training loop is to add DL parts to it.

All folder architecture + envs is done