Update requirements.txt and add working custom env code (from README)

blokus_gym/__init__.py

@@ -1,5 +1,10 @@
 from gym.envs.registration import register
 
+register(
+    id='blokus-custom-v0',
+    entry_point='blokus_gym.envs:BlokusCustomEnv',
+)
+
 register(
     id='blokus-hard-v0',
     entry_point='blokus_gym.envs:BlokusEnv',

blokus_gym/envs/__init__.py

@@ -1,2 +1,2 @@
 from blokus_gym.envs.blokus_env import BlokusEnv
-from blokus_gym.envs.blokus_envs import BlokusSimpleEnv, BlokusSimpleGreedyEnv, BlokusDuoEnv, BlokusDuoGreedyEnv, BlokusGreedyEnv, BlokusSimpleMinimaxEnv
+from blokus_gym.envs.blokus_envs import BlokusSimpleEnv, BlokusSimpleGreedyEnv, BlokusDuoEnv, BlokusDuoGreedyEnv, BlokusGreedyEnv, BlokusSimpleMinimaxEnv, BlokusCustomEnv

blokus_gym/envs/blokus_envs.py

@@ -2,7 +2,15 @@
 from blokus_gym.envs.shapes.shapes import get_all_shapes
 from blokus_gym.envs.players.greedy_player import GreedyPlayer
 from blokus_gym.envs.players.minimax_player import MinimaxPlayer
-
+from blokus_gym.envs.players.random_player import RandomPlayer
+
+class BlokusCustomEnv(BlokusEnv):
+    NUMBER_OF_PLAYERS = 3
+    BOARD_SIZE = 10 # This will result in a 10x10 board
+    STATES_FILE = "states.json"  # This needs to be set, if not it will take the base class states
+    all_shapes = [shape for shape in get_all_shapes()
+                  if shape.size == 4]  # This will take only the 4 tiles pieces
+    bot_type = RandomPlayer  # Defaults to RandomPlayer if not passed
 
 class BlokusGreedyEnv(BlokusEnv):
     bot_type = GreedyPlayer

play_custom.py

@@ -0,0 +1,35 @@
+import blokus_gym
+from blokus_gym.envs.blokus_env import BlokusEnv
+import gym
+import random
+
+if __name__ == "__main__":
+    # env = BlokusEnv()
+    env = gym.make("blokus_gym:blokus-custom-v0")  # Make sure to do: pip install -e . in root
+    print(f"number of possible moves {env.action_space}")
+    count = 0
+    nb_rounds = 0
+    for _ in range(100):
+        while True:
+            # input()
+            action = env.action_space.sample()
+            count += len(env.ai_possible_indexes())
+            nb_rounds += 1
+            # action = random.randint(0, 918)
+            observation, reward, done, info = env.step(action)
+            env.render("human")
+            # print(env.ai.all_ids_to_move.keys())
+            # print(reward)
+
+            if done:
+                # print(env.ai.all_ids_to_move.keys())
+                # input()
+                print(f"{'won' if reward == 1 else ('tie-won' if reward == 0 else 'lost')}")
+                # print(env.ai.all_ids_to_move.keys())
+                observation = env.reset()
+                # input()
+                break
+
+    print(f"Average number of moves per turn: {count / nb_rounds:.2f}")
+    print(f"Starter won {env.starter_won / env.games_played * 100:.2f}%")
+    env.close()

requirements.txt

@@ -1,5 +1,10 @@
-numpy>=1.17.4
-gym>=0.17.1
-torch>=1.4.0
+cython>=0.29.16
+gym==0.15.7
 matplotlib>=3.1.2
-cython>=0.29.16
+numpy>=1.17.4
+opencv-python==4.5.3.56
+pandas>=1.1.5
+setuptools>=58.0.4
+stable_baselines>=2.10.2
+tensorflow==1.15.2
+torch>=1.4.0