Merge pull request #630 from DEUCE1957/EpisodeData

Debug: Fix Missing Legal and Ambigious Arrays in CompactEpisodeData

Author: BDonnot

grid2op/Episode/CompactEpisodeData.py

@@ -95,9 +95,11 @@ def __init__(self, env, obs, exp_dir, ep_id:str=None):
         """
         if exp_dir is not None:
             self.exp_dir = p(exp_dir)
+            self.exp_dir = self.exp_dir / "CompactEpisodeData"
+            self.exp_dir.mkdir(parents=False, exist_ok=True)
         else:
             self.exp_dir = None
-        self.array_names = ("actions", "env_actions", "attacks", "observations", "rewards", "other_rewards", "disc_lines", "times")
+        self.array_names = ("actions", "env_actions", "attacks", "observations", "rewards", "other_rewards", "disc_lines", "times", "legal", "ambiguous")
         self.space_names = ("observation_space", "action_space", "attack_space", "env_modification_space")
         if ep_id is None:
             self.ep_id = env.chronics_handler.get_name()

grid2op/Runner/runner.py

@@ -1728,7 +1728,7 @@ def run(
                     )
                 else:
                     if add_detailed_output and (_IS_WINDOWS or _IS_MACOS):
-                        self.logger.warn(
+                        self.logger.warning(
                             "Parallel run are not fully supported on windows or macos when "
                             '"add_detailed_output" is True. So we decided '
                             "to fully deactivate them."

grid2op/tests/test_CompactEpisodeData.py

@@ -12,7 +12,7 @@
 import unittest
 
 import grid2op
-from grid2op.Agent import OneChangeThenNothing
+from grid2op.Agent import DoNothingAgent, OneChangeThenNothing
 from grid2op.tests.helper_path_test import *
 from grid2op.Chronics import Multifolder
 from grid2op.Reward import L2RPNReward
@@ -140,6 +140,8 @@ def act(self, observation, reward, done=False):
         assert len(episode_data.observations) == self.max_iter + 1
         assert len(episode_data.env_actions) == self.max_iter
         assert len(episode_data.attacks) == self.max_iter
+        assert len(episode_data.ambiguous) == self.max_iter
+        assert len(episode_data.legal) == self.max_iter
 
     def test_one_episode_with_saving(self):
         f = tempfile.mkdtemp()
@@ -163,6 +165,7 @@ def test_collection_wrapper_after_run(self):
         OneChange = OneChangeThenNothing.gen_next(
             {"set_bus": {"lines_or_id": [(1, -1)]}}
         )
+        # env.reset(options=)
         runner = Runner(
             init_grid_path=self.init_grid_path,
             init_env_path=self.init_grid_path,
@@ -178,9 +181,11 @@ def test_collection_wrapper_after_run(self):
             agentClass=OneChange,
             use_compact_episode_data=True,
         )
-        ep_id, ep_name, cum_reward, timestep, max_ts, episode_data = runner.run_one_episode(
-            max_iter=self.max_iter, detailed_output=True
-        )
+        with warnings.catch_warnings():
+            warnings.filterwarnings("ignore")
+            *_, episode_data = runner.run_one_episode(
+                max_iter=self.max_iter, detailed_output=True,
+            )
         # Check that the type of first action is set bus
         assert episode_data.action_space.from_vect(episode_data.actions[0]).get_types()[2]
 
@@ -257,7 +262,7 @@ def test_with_opponent(self):
         )
 
         episode_data = CompactEpisodeData.from_disk(path=f, ep_id=res[0][1])
-        lines_impacted, subs_impacted = episode_data.attack_space.from_vect(episode_data.attacks[0]).get_topological_impact()
+        lines_impacted, _ = episode_data.attack_space.from_vect(episode_data.attacks[0]).get_topological_impact()
         assert lines_impacted[3]
 
     def test_can_return_ep_data(self):
@@ -296,4 +301,4 @@ def test_can_return_ep_data(self):
 
 
 if __name__ == "__main__":
-    unittest.main()
+    unittest.main()