fix indentation and issues with shape.

PokerPlus/DeepCFR/game_tree.py

@@ -17,7 +17,7 @@
 
 from PokerPlus.DeepCFR.memory import AdvantageMemory, StrategyMemory
 
-from PokerPlus.DeepCFR.utils import get_opponent_player_num
+from PokerPlus.DeepCFR.utils import get_opponent_player_num, available_moves
 
 
 def compute_strategy(state: TexasHoldEm, strategy_net: DeepCFRModel, nb_actions=5):
@@ -28,13 +28,22 @@ def compute_strategy(state: TexasHoldEm, strategy_net: DeepCFRModel, nb_actions=
     """
 
     # Get the available moves for the current player
-    legal_actions = state.get_available_moves()[:nb_actions]
+    print("state: ", state)
+    legal_actions = available_moves(state, nb_actions)
     player = state.current_player
 
     cards_board = [card_to_int[str(card)] for card in state.board]
+    flop = [card for card in cards_board[:3]]
+    turn = cards_board[3:4]
+    river = cards_board[4:5]
     hole = state.get_hand(player)
     hole = [card_to_int[str(card)] for card in hole]
-    cards = hole + cards_board
+    cards = [hole, flop]
+    if turn:
+        cards.extend([turn])
+
+    if river:
+        cards.extend([river])
     # bets
     bets = [val_bet for val_bet in state._get_last_pot().player_amounts.values()]
     cards = [torch.tensor(c, dtype=torch.long) for c in cards]
@@ -73,17 +82,6 @@ def get_info_set(game: TexasHoldEm, player: int):
     return cards, bets
 
 
-def available_moves(game : TexasHoldEm, nb_actions : int):
-    actions = game.get_available_moves()
-    available_actions = []
-    for index, action in enumerate(actions):
-        if action[0] < nb_actions:
-            available_actions.append(action[1])
-        else:
-            return available_actions
-
-
-
 def traverse(
     game: TexasHoldEm,
     current_player_to_compute_strategy: int,
@@ -128,9 +126,10 @@ def traverse(
         )  # Insert infoset and its action advantages
         return np.max(action_values)  # Return the value of the best action
     else:
+        available_moves_ = available_moves(game, nb_actions)
         opponent_num = get_opponent_player_num(current_player_to_compute_strategy)
         sigma_t = compute_strategy(
-            get_info_set(game, opponent_num), theta2, nb_actions
+            game, theta2, nb_actions
         )  # Compute opponent's strategy
 
         StrategyMemory.insert(

PokerPlus/DeepCFR/nn.py

@@ -20,6 +20,8 @@ def __init__(self, dim):
         self.card = nn.Embedding(52, dim)
 
     def forward(self, input):
+        print("forward of CardEmbedding")
+        print("input: ", input)
         B, num_cards = input.shape
         x = input.view(-1)
         valid = x.ge(0).float()  # -1 means 'no card'
@@ -54,8 +56,13 @@ def forward(self, cards, bets):
         # 1. card branch
         # embed hole, flop, and optionally turn and river
         card_embs = []
+        print("cards: ", cards)
+        print("bets: ", bets)
+
         for embedding, card_group in zip(self.card_embeddings, cards):
-            card_embs.append(embedding(card_group))
+            print("do embedding")
+            print("card_group: ", card_group)
+            card_embs.append(embedding(card_group.view(1, -1)))
         card_embs = torch.cat(card_embs, dim=1)
         x = F.relu(self.card1(card_embs))
         x = F.relu(self.card2(x))

PokerPlus/DeepCFR/utils.py

@@ -1,5 +1,6 @@
 from texasholdem.card.deck import Deck
 from texasholdem.card.card import Card
+from texasholdem.game.game import TexasHoldEm
 
 d = Deck()
 card_to_int = {}
@@ -17,4 +18,14 @@ def get_opponent_player_num(current_player: int):
     if current_player == 1:
         return 0
     else:
-        return 1
+        return 1
+
+
+def available_moves(game: TexasHoldEm, nb_actions: int):
+    actions = game.get_available_moves()
+    available_actions = []
+    for index, action in enumerate(actions):
+        if index < nb_actions:
+            available_actions.append(action[1])
+        else:
+            return available_actions

random_test.py

@@ -13,5 +13,5 @@
         print("game available moves slicing : ", game.get_available_moves()[:5])
         for move in game.get_available_moves():
             print("move : ", move)
-            
-        gui.run_step()
+
+        gui.run_step()

test.py

@@ -1,12 +1,23 @@
 from PokerPlus.DeepCFR.deep_cfr import deep_cfr, save_deep_cfr
 from texasholdem.game.game import TexasHoldEm
 
+
 def main():
     print("create game")
     game = TexasHoldEm(buyin=1500, big_blind=80, small_blind=40, max_players=2)
     game.start_hand()
-    save_deep_cfr("", "DeepCFR", nb_iterations=10000, nb_players=2, nb_game_tree_traversals=200, game=game, n_actions=3, n_card_types=4, n_bets=20)
+    save_deep_cfr(
+        "",
+        "DeepCFR",
+        nb_iterations=10000,
+        nb_players=2,
+        nb_game_tree_traversals=200,
+        game=game,
+        n_actions=3,
+        n_card_types=4,
+        n_bets=20,
+    )
 
 
 if __name__ == "__main__":
-    main()
+    main()