dodox/mujoco_maze
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Some style fixes

Browse Source
This commit is contained in:
kngwyu 2020-07-01 14:20:23 +09:00
parent 14351830bb
commit 4001fef23b
1 changed files with 3 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
mujoco_maze/maze_task.py
View File

@ -51,7 +51,7 @@ class Scaling(NamedTuple):
class MazeTask(ABC): class MazeTask(ABC):
REWARD_THRESHOLD: float REWARD_THRESHOLD: float
MAZE_SIZE_SCALING: Scaling = Scaling(8.0, 4.0) MAZE_SIZE_SCALING: Scaling = Scaling(8.0, 4.0)
INNER_REWARD_SCALING: float = 1e-4 INNER_REWARD_SCALING: float = 0.0
OBSERVE_BLOCKS: bool = False OBSERVE_BLOCKS: bool = False
PUT_SPIN_NEAR_AGENT: bool = False PUT_SPIN_NEAR_AGENT: bool = False
@ -103,7 +103,6 @@ class SingleGoalSparseUMaze(MazeTask):
class SingleGoalDenseUMaze(SingleGoalSparseUMaze): class SingleGoalDenseUMaze(SingleGoalSparseUMaze):
REWARD_THRESHOLD: float = 1000.0 REWARD_THRESHOLD: float = 1000.0
def reward(self, obs: np.ndarray) -> float: def reward(self, obs: np.ndarray) -> float:
return -self.goals[0].euc_dist(obs) return -self.goals[0].euc_dist(obs)
@ -128,7 +127,6 @@ class SingleGoalSparsePush(SingleGoalSparseUMaze):
class SingleGoalDensePush(SingleGoalSparsePush): class SingleGoalDensePush(SingleGoalSparsePush):
REWARD_THRESHOLD: float = 1000.0 REWARD_THRESHOLD: float = 1000.0
def reward(self, obs: np.ndarray) -> float: def reward(self, obs: np.ndarray) -> float:
return -self.goals[0].euc_dist(obs) return -self.goals[0].euc_dist(obs)
@ -154,14 +152,13 @@ class SingleGoalSparseFall(SingleGoalSparseUMaze):
class SingleGoalDenseFall(SingleGoalSparseFall): class SingleGoalDenseFall(SingleGoalSparseFall):
REWARD_THRESHOLD: float = 1000.0 REWARD_THRESHOLD: float = 1000.0
def reward(self, obs: np.ndarray) -> float: def reward(self, obs: np.ndarray) -> float:
return -self.goals[0].euc_dist(obs) return -self.goals[0].euc_dist(obs)
class SingleGoalSparse2Rooms(MazeTask): class SingleGoalSparse2Rooms(MazeTask):
REWARD_THRESHOLD: float = 0.9 REWARD_THRESHOLD: float = 0.9
SCALING: Scaling = Scaling(4.0, 4.0) MAZE_SIZE_SCALING: Scaling = Scaling(4.0, 4.0)
def __init__(self, scale: float) -> None: def __init__(self, scale: float) -> None:
super().__init__(scale) super().__init__(scale)
@ -190,7 +187,6 @@ class SingleGoalSparse2Rooms(MazeTask):
class SingleGoalDense2Rooms(SingleGoalSparse2Rooms): class SingleGoalDense2Rooms(SingleGoalSparse2Rooms):
REWARD_THRESHOLD: float = 1000.0 REWARD_THRESHOLD: float = 1000.0
def reward(self, obs: np.ndarray) -> float: def reward(self, obs: np.ndarray) -> float:
return -self.goals[0].euc_dist(obs) return -self.goals[0].euc_dist(obs)
@ -234,7 +230,6 @@ class SingleGoalSparse4Rooms(MazeTask):
class SingleGoalDense4Rooms(SingleGoalSparse4Rooms): class SingleGoalDense4Rooms(SingleGoalSparse4Rooms):
REWARD_THRESHOLD: float = 1000.0 REWARD_THRESHOLD: float = 1000.0
def reward(self, obs: np.ndarray) -> float: def reward(self, obs: np.ndarray) -> float:
return -self.goals[0].euc_dist(obs) return -self.goals[0].euc_dist(obs)
@ -253,7 +248,7 @@ class TaskRegistry:
"UMaze": [SingleGoalDenseUMaze, SingleGoalSparseUMaze], "UMaze": [SingleGoalDenseUMaze, SingleGoalSparseUMaze],
"Push": [SingleGoalDensePush, SingleGoalSparsePush], "Push": [SingleGoalDensePush, SingleGoalSparsePush],
"Fall": [SingleGoalDenseFall, SingleGoalSparseFall], "Fall": [SingleGoalDenseFall, SingleGoalSparseFall],
"2Rooms": [SingleGoalDense2Rooms, SingleGoalSparse2Rooms, SubGoalSparse2Rooms,], "2Rooms": [SingleGoalDense2Rooms, SingleGoalSparse2Rooms, SubGoalSparse2Rooms],
"4Rooms": [SingleGoalSparse4Rooms, SingleGoalDense4Rooms, SubGoalSparse4Rooms], "4Rooms": [SingleGoalSparse4Rooms, SingleGoalDense4Rooms, SubGoalSparse4Rooms],
} }