gaspersic/Gymnasium
1
0
Fork 0
mirror of https://github.com/Farama-Foundation/Gymnasium.git synced 2025-08-31 18:12:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix victory conditions

Browse Source 
Old victory condition was "player is white and white wins." Added "player is black and black wins" and "opponent resigned."
This commit is contained in:
aforr
2016-04-28 20:30:50 -04:00
parent 100bc82e38
commit c8ada04b34
1 changed files with 5 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
gym/envs/board_game/go.py
View File

@@ -201,7 +201,7 @@ class GoEnv(gym.Env):
# Opponent play # Opponent play
if not self.state.board.is_terminal: if not self.state.board.is_terminal:
self.state = self._exec_opponent_play(self.state, prev_state, action) self.state, self.opponent_resigned = self._exec_opponent_play(self.state, prev_state, action)
# After opponent play, we should be back to the original color # After opponent play, we should be back to the original color
assert self.state.color == self.player_color assert self.state.color == self.player_color
@@ -209,7 +209,8 @@ class GoEnv(gym.Env):
if self.state.board.is_terminal: if self.state.board.is_terminal:
self.done = True self.done = True
white_wins = self.state.board.official_score > 0 white_wins = self.state.board.official_score > 0
reward = 1. if (white_wins and self.player_color == pachi_py.WHITE) else -1. player_wins = (white_wins and self.player_color == pachi_py.WHITE) or (not white_wins and self.player_color == pachi_py.BLACK)
reward = 1. if (self.opponent_resigned or player_wins) else -1.
else: else:
self.done = False self.done = False
reward = 0. reward = 0.
@@ -218,7 +219,8 @@ class GoEnv(gym.Env):
def _exec_opponent_play(self, curr_state, prev_state, prev_action): def _exec_opponent_play(self, curr_state, prev_state, prev_action):
assert curr_state.color != self.player_color assert curr_state.color != self.player_color
opponent_action = self.opponent_policy(curr_state, prev_state, prev_action) opponent_action = self.opponent_policy(curr_state, prev_state, prev_action)
return curr_state.act(opponent_action) opponent_resigned = opponent_action == self.board_size**2+1
return curr_state.act(opponent_action), opponent_resigned
@property @property
def _state(self): def _state(self):