1

1

scripts/train_chengdu_mahjong_model.py

@@ -1,22 +1,28 @@
 import gym
 from stable_baselines3 import PPO
-from src.environment.chengdu_mahjong_env import MahjongEnv
+from src.environment.chengdu_mahjong_env import ChengduMahjongEnv
 import torch
 from configs.log_config import setup_logging
 
 def train_model():
     # 创建 MahjongEnv 环境实例
-    env = MahjongEnv()
+    env = ChengduMahjongEnv()
 
     # 检查是否有可用的 GPU
     device = "cuda" if torch.cuda.is_available() else "cpu"
     print(f"使用设备: {device}")
 
-    # 使用 PPO 算法训练模型
+    # 使用 PPO 算法训练模型，切换到 MultiInputPolicy
-    model = PPO("MlpPolicy", env, verbose=1, tensorboard_log="../logs/ppo_mahjong_tensorboard/", device=device)
+    model = PPO(
+        "MultiInputPolicy",  # 更改为 MultiInputPolicy
+        env,
+        verbose=1,
+        tensorboard_log="../logs/ppo_mahjong_tensorboard/",
+        device=device
+    )
 
     # 训练模型，训练总步数为 100000
-    model.learn(total_timesteps=100)
+    model.learn(total_timesteps=100000)
 
     # 保存训练后的模型
     model.save("../models/ppo_mahjong_model")
@@ -27,7 +33,7 @@ def train_model():
     while not done:
         action, _states = model.predict(obs)  # 使用训练好的模型来选择动作
         obs, reward, done, info = env.step(action)  # 执行动作
-        env.render()  # 打印环境状态
+        print(f"动作: {action}, 奖励: {reward}, 是否结束: {done}, 信息: {info}")
 
 if __name__ == "__main__":
     # 调用配置函数来设置日志

src/engine/actions.py

@@ -4,6 +4,7 @@ from loguru import logger
 
 from src.engine.calculate_fan import calculate_fan
 from src.engine.mahjong_tile import MahjongTile
+from src.engine.meld import Meld
 
 
 def draw_tile(engine):
@@ -235,15 +236,16 @@ def handle_peng(self, player, tile):
     处理玩家碰牌逻辑并更新出牌顺序。
     """
     if not isinstance(tile, MahjongTile):
-        logger.error(f"玩家 {player} 碰牌的牌无效: {tile}")
+        logger.error(f"tile 必须是 MahjongTile 类型，但收到的是: {type(tile)}")
         return False
 
     if self.state.hands[player].tile_count[tile] < 2:
         logger.error(f"玩家 {player} 无法碰牌: {tile}")
         return False
 
-    # 更新状态
+    # 减少两张牌
-    self._update_meld(player, tile, "碰", count=2)
+    self.state.hands[player].tile_count[tile] -= 2
+    self.state.melds[player].append(Meld(tile, "碰"))  # 确保使用 MahjongTile 对象
 
     logger.info(f"玩家 {player} 碰了牌: {tile}。当前明牌: {self.state.melds[player]}")
     return True
@@ -251,6 +253,8 @@ def handle_peng(self, player, tile):
 
 
 
+
+
 def get_player_discard_choice(self, player):
     """
     模拟获取玩家打牌的选择。

src/engine/mahjong_tile.py

@@ -6,6 +6,7 @@ class MahjongTile:
             raise ValueError("Invalid tile")
         self.suit = suit
         self.value = value
+        self.index = ({"条": 0, "筒": 1, "万": 2}[suit]) * 9 + (value - 1)
 
     def __repr__(self):
         return f"{self.value}{self.suit}"

src/environment/chengdu_mahjong_env.py

@@ -1,6 +1,8 @@
 import gym
 from gym import spaces
 import numpy as np
+
+from src.engine.actions import handle_peng, handle_gang, handle_win
 from src.engine.chengdu_mahjong_engine import ChengduMahjongEngine
 from loguru import logger
 
@@ -50,13 +52,13 @@ class ChengduMahjongEnv(gym.Env):
         elif action == 14:  # 碰
             tile_to_peng = self._get_tile_for_special_action("peng")
             if tile_to_peng:
-                self.engine.handle_peng(current_player, tile_to_peng)
+                handle_peng(self.engine,current_player, tile_to_peng)
             else:
                 logger.warning("碰动作无效，未满足条件")
         elif action == 15:  # 杠
             tile_to_gang = self._get_tile_for_special_action("gang")
             if tile_to_gang:
-                self.engine.handle_gang(current_player, tile_to_gang, mode="an")  # 默认暗杠
+                handle_gang(current_player, tile_to_gang, mode="an")  # 默认暗杠
             else:
                 logger.warning("杠动作无效，未满足条件")
         elif action == 16:  # 胡
@@ -65,7 +67,7 @@ class ChengduMahjongEnv(gym.Env):
                 self.engine.state.melds[current_player],
                 self.engine.state.missing_suits[current_player]
             ):
-                self.engine.handle_win(current_player, None, None)
+                handle_win(current_player, None, None)
             else:
                 logger.warning("胡动作无效，未满足条件")
         else: