Fix Poolfiz errors

P6-Pool · May 13, 2024 · 4abfc75 · 4abfc75
1 parent 3c8e941
commit 4abfc75
Show file tree

Hide file tree

Showing 4 changed files with 12 additions and 16 deletions.
diff --git a/src/fastfiz_env/envs/pockets_fastfiz.py b/src/fastfiz_env/envs/pockets_fastfiz.py
@@ -80,10 +80,7 @@ def step(self, action: np.ndarray) -> tuple[np.ndarray, float, bool, bool, dict]
         shot_params = ff.ShotParams(*action)
 
         if self._possible_shot(shot_params):
-            try:
-                self.table_state.executeShot(shot_params)
-            except Exception:
-                pass
+            self.table_state.executeShot(shot_params, verbose=True)
 
         observation = self._get_observation()
         reward = self.reward.get_reward(prev_table_state, self.table_state, action)
@@ -106,11 +103,11 @@ def compute_observation(cls, table_state: ff.TableState) -> np.ndarray:
             if ball.isPocketed():
                 pocket = ball_state_to_pocket(ball.getState())
                 pocket_pos = get_pocket_center(pocket)
-                observation[i] = [*pocket_pos, 0]
+                observation[i] = [*pocket_pos, 1]
             elif ball.isInPlay():
-                observation[i] = [*ball_pos, 1]
+                observation[i] = [*ball_pos, 0]
             else:
-                observation[i] = [0, 0, 0]
+                observation[i] = [0, 0, 1]
 
         return np.array(observation)
 

diff --git a/src/fastfiz_env/envs/simple_fastfiz.py b/src/fastfiz_env/envs/simple_fastfiz.py
@@ -76,10 +76,7 @@ def step(self, action: np.ndarray) -> tuple[np.ndarray, float, bool, bool, dict]
         shot_params = ff.ShotParams(*action)
 
         if self._possible_shot(shot_params):
-            try:
-                self.table_state.executeShot(shot_params)
-            except Exception:
-                pass
+            self.table_state.executeShot(shot_params)
 
         observation = self._get_observation()
         reward = self.reward.get_reward(prev_table_state, self.table_state, action)

diff --git a/src/fastfiz_env/wrappers/action.py b/src/fastfiz_env/wrappers/action.py
@@ -72,8 +72,8 @@ class FastFizActionWrapper(ActionWrapper):
     MAX_PHI = 360 - 0.001
     MIN_VELOCITY = 0
     MAX_VELOCITY = 10 - 0.001
-    MIN_OFFSET = -28
-    MAX_OFFSET = 28
+    MIN_OFFSET = -28 + 0.001
+    MAX_OFFSET = 28 - 0.001
     SPACES = {
         "VECTOR_2D": spaces.Box(
             low=np.array([-1, -1]),

diff --git a/src/train.py b/src/train.py
@@ -50,6 +50,8 @@ def train(
         n_envs=n_envs,
     )
 
+    # env = VecCheckNan(env, raise_exception=True)
+
     hyperparams = params_to_kwargs(**params) if params else {}
     print(hyperparams)
     model_name = get_model_name(env_id, num_balls, action_space_id=action_space_id)
@@ -97,8 +99,8 @@ def train(
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
-    parser.add_argument("--env", type=str, required=True)
-    parser.add_argument("-b", "--num_balls", type=int, required=True)
+    parser.add_argument("--env", type=str, default="PocketsFastFiz-v0")
+    parser.add_argument("-b", "--num_balls", type=int, default=2)
     parser.add_argument("-m", "--max_episode_steps", type=int, default=20)
     parser.add_argument("-n", "--n_time_steps", type=int, default=1_000_000)
     parser.add_argument(
@@ -125,7 +127,7 @@ def train(
     )
 
     parser.add_argument(
-        "-a", "--action_id", type=lambda a: ActionSpaces[a], choices=list(ActionSpaces), default=ActionSpaces.VECTOR_3D
+        "-a", "--action_id", type=lambda a: ActionSpaces[a], choices=list(ActionSpaces), default=ActionSpaces.OFFSET_NORM_5D
     )
 
     args = parser.parse_args()