ver 2.1.0 ready

humemai · Sep 25, 2024 · 77ec4ea · 77ec4ea
1 parent bca245d
commit 77ec4ea
Show file tree

Hide file tree

Showing 7 changed files with 52 additions and 51 deletions.
diff --git a/README.md b/README.md
@@ -23,7 +23,6 @@ repo](https://github.com/humemai/humemai).
 - ["A Machine with Short-Term, Episodic, and Semantic Memory
   Systems"](https://arxiv.org/abs/2212.02098)
 - ["Leveraging Knowledge Graph-Based Human-Like Memory Systems to Solve Partially Observable Markov Decision Processes"](https://arxiv.org/abs/2408.05861)
-
 
 ## pdoc documentation
 

diff --git a/room_env/envs/room2.py b/room_env/envs/room2.py
@@ -839,7 +839,7 @@ def step(self, actions: tuple[list[str], str]) -> tuple[dict, int, bool, dict]:
                     north, east, south, west, or stay.
 
         Returns:
-            (observation, question), reward, truncated, done, info
+            (observation, question), rewards, truncated, done, info
 
         """
         actions_qa, action_explore = actions
@@ -849,20 +849,20 @@ def step(self, actions: tuple[list[str], str]) -> tuple[dict, int, bool, dict]:
 
         if len(self.answers) == 0:
             assert actions_qa == [], "You shouldn't answer any questions"
-            reward = 0
+            rewards = []
 
         else:
             assert len(actions_qa) == len(
                 self.answers
             ), "You should answer all the questions."
-            reward = 0
+            rewards = []
             for actions_qa, answer in zip(actions_qa, self.answers):
                 if actions_qa == answer["current"]:
-                    reward += self.CORRECT
+                    rewards.append(self.CORRECT)
                 elif actions_qa == answer["previous"]:
-                    reward += self.PARTIAL
+                    rewards.append(self.PARTIAL)
                 else:
-                    reward += self.WRONG
+                    rewards.append(self.WRONG)
 
         if not self.make_everything_static:
             for obj in self.objects["independent"]:
@@ -891,15 +891,15 @@ def step(self, actions: tuple[list[str], str]) -> tuple[dict, int, bool, dict]:
         if (self.current_time + 1) % self.question_interval == 0:
             return (
                 self.get_observations_and_question(generate_questions=True),
-                reward,
+                rewards,
                 done,
                 truncated,
                 info,
             )
         else:
             return (
                 self.get_observations_and_question(generate_questions=False),
-                reward,
+                rewards,
                 done,
                 truncated,
                 info,

diff --git a/setup.cfg b/setup.cfg
@@ -1,6 +1,6 @@
 [metadata]
 name = room_env
-version = 2.0.7
+version = 2.1.0
 author = Taewoon Kim
 author_email = info@humem.ai
 description = The Room environment

diff --git a/test/room_env2/test_history.py b/test/room_env2/test_history.py
@@ -45,7 +45,7 @@ def test_all_static(self) -> None:
                     locations[obj.name] = [obj.location]
 
         while True:
-            observations, reward, done, truncated, info = env.step((["foo"], "stay"))
+            observations, rewards, done, truncated, info = env.step((["foo"], "stay"))
             for obj_type, objs in env.objects.items():
                 for obj in objs:
                     locations[obj.name].append(obj.location)
@@ -80,7 +80,7 @@ def test_not_all_static(self) -> None:
                     locations[obj.name] = [obj.location]
 
         while True:
-            observations, reward, done, truncated, info = env.step(
+            observations, rewards, done, truncated, info = env.step(
                 (["foo"], random.choice(["north", "south", "east", "west"]))
             )
             for obj_type, objs in env.objects.items():
@@ -122,16 +122,16 @@ def test_deterministic_objects(self) -> None:
                 }
                 env = gym.make("room_env:RoomEnv-v2", **env_config)
                 observations, info = env.reset()
-                rewards = 0
+                rewards = []
 
                 while True:
-                    observations, reward, done, truncated, info = env.step(
+                    observations, rewards_, done, truncated, info = env.step(
                         (
                             ["random answer"] * len(observations["questions"]),
                             random.choice(["north", "east", "south", "west", "stay"]),
                         )
                     )
-                    rewards += reward
+                    rewards.append(rewards_)
                     if done or truncated:
                         break
 

diff --git a/test/room_env2/test_num_total_questions_question_interval.py b/test/room_env2/test_num_total_questions_question_interval.py
@@ -42,9 +42,9 @@ def test_wrong_answers(self) -> None:
 
         self.assertEqual(observations["questions"], [])
         with self.assertRaises(AssertionError):
-            observations, reward, done, truncated, info = env.step(("foo", "stay"))
+            observations, rewards, done, truncated, info = env.step(("foo", "stay"))
         with self.assertRaises(AssertionError):
-            observations, reward, done, truncated, info = env.step((["foo"], "stay"))
+            observations, rewards, done, truncated, info = env.step((["foo"], "stay"))
 
     def test_correct_answers(self) -> None:
         env_config = {
@@ -64,8 +64,8 @@ def test_correct_answers(self) -> None:
         questions_all = []
         self.assertEqual(len(observations["questions"]), 0)
 
-        observations, reward, done, truncated, info = env.step(([], "stay"))
-        self.assertEqual(reward, 0)
+        observations, rewards, done, truncated, info = env.step(([], "stay"))
+        self.assertEqual(rewards, [])
 
         while True:
             flag = False
@@ -77,11 +77,13 @@ def test_correct_answers(self) -> None:
                 for q in observations["questions"]:
                     questions_all.append(q)
 
-            observations, reward, done, truncated, info = env.step((actions_qa, "stay"))
+            observations, rewards, done, truncated, info = env.step(
+                (actions_qa, "stay")
+            )
             if flag:
-                self.assertEqual(reward, -1 * env.num_questions_step)
+                self.assertEqual(rewards, [-1] * env.num_questions_step)
             else:
-                self.assertEqual(reward, 0)
+                self.assertEqual(rewards, [])
 
             if done:
                 break

diff --git a/test/room_env2/test_partial_points.py b/test/room_env2/test_partial_points.py
@@ -33,26 +33,26 @@ def test_partial_points(self) -> None:
                 if break_:
                     break
             if len(set(obj.history)) == 1:
-                observations, reward, done, truncated, info = env.step(
+                observations, rewards, done, truncated, info = env.step(
                     (
                         [obj.location],
                         random.choice(["north", "south", "east", "west", "stay"]),
                     )
                 )
 
-                self.assertEqual(reward, 1)
+                self.assertEqual(rewards, [1])
             else:
                 for previous_location in obj.history[::-1]:
                     if previous_location != obj.location:
                         break
 
-                observations, reward, done, truncated, info = env.step(
+                observations, rewards, done, truncated, info = env.step(
                     (
                         [previous_location],
                         random.choice(["north", "south", "east", "west", "stay"]),
                     )
                 )
-                self.assertEqual(reward, 0)
+                self.assertEqual(rewards, [0])
 
             if done:
                 break
diff --git a/test/room_env2/test_room_env_v2.py b/test/room_env2/test_room_env_v2.py
@@ -88,11 +88,11 @@ def test_all(self) -> None:
 
             question_previous = observations["questions"][0]
 
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (actions_qa, "east")
             )
-            rewards.append(reward)
-            self.assertEqual(reward, 1)
+            rewards.append(rewards_)
+            self.assertEqual(rewards_, [1])
             self.assertFalse(done)
             # if question_previous == ["?", "atlocation", "officeroom", 0]:
             #     self.assertEqual(info, {"answers": ["desk"], "timestamp": 0})
@@ -131,11 +131,11 @@ def test_all(self) -> None:
                 raise ValueError
             question_previous = observations["questions"][0]
 
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (actions_qa, "stay")
             )
-            rewards.append(reward)
-            self.assertEqual(reward, 1)
+            rewards.append(rewards_)
+            self.assertEqual(rewards_, [1])
             self.assertFalse(done)
             # if question_previous == ["?", "atlocation", "officeroom", 1]:
             #     self.assertEqual(info, {"answers": ["desk"], "timestamp": 1})
@@ -151,20 +151,20 @@ def test_all(self) -> None:
                 raise ValueError
 
             for _ in range(7):
-                observations, reward, done, truncated, info = self.env.step(
+                observations, rewards_, done, truncated, info = self.env.step(
                     (["foo"], "stay")
                 )
-                rewards.append(reward)
+                rewards.append(rewards_)
 
-                self.assertEqual(reward, -1)
+                self.assertEqual(rewards_, [-1])
                 self.assertFalse(done)
 
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (["bar"], "stay")
             )
-            rewards.append(reward)
+            rewards.append(rewards_)
             self.assertEqual(len(rewards), 10)
-            self.assertEqual(reward, -1)
+            self.assertEqual(rewards_, [-1])
             self.assertTrue(done)
             # self.assertIsNone(self.env.observations_room)
             # self.assertIsNone(self.env.question)
@@ -296,10 +296,10 @@ def test_all(self) -> None:
                 actions_qa = ["officeroom"]
             question_previous = observations["questions"]
 
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (actions_qa, "east")
             )
-            rewards.append(reward)
+            rewards.append(rewards_)
             # if question_previous[0] == "?":
             #     self.assertEqual(
             #         info, {"answers": ["desk", "tae", "laptop"], "timestamp": 0}
@@ -335,7 +335,7 @@ def test_all(self) -> None:
                     ["?", "atlocation", "livingroom", 1],
                 ],
             )
-            self.assertEqual(reward, 1)
+            self.assertEqual(rewards_, [1])
             self.assertFalse(done)
 
             if observations["questions"][0] == ["desk", "atlocation", "?", 1]:
@@ -353,10 +353,10 @@ def test_all(self) -> None:
 
             question_previous = observations["questions"][0]
 
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (actions_qa, "west")
             )
-            rewards.append(reward)
+            rewards.append(rewards_)
             if question_previous == ["desk", "atlocation", "?", 1]:
                 self.assertEqual(
                     info,
@@ -416,24 +416,24 @@ def test_all(self) -> None:
                     ["?", "atlocation", "officeroom", 2],
                 ],
             )
-            self.assertEqual(reward, 1)
+            self.assertEqual(rewards_, [1])
             self.assertFalse(done)
 
             for _ in range(97):
-                observations, reward, done, truncated, info = self.env.step(
+                observations, rewards_, done, truncated, info = self.env.step(
                     (["foo"], "stay")
                 )
-                rewards.append(reward)
+                rewards.append(rewards_)
 
-                self.assertEqual(reward, -1)
+                self.assertEqual(rewards_, [-1])
                 self.assertFalse(done)
 
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (["bar"], "stay")
             )
-            rewards.append(reward)
+            rewards.append(rewards_)
             self.assertEqual(len(rewards), 100)
-            self.assertEqual(reward, -1)
+            self.assertEqual(rewards_, [-1])
             self.assertTrue(done)
             # self.assertIsNone(self.env.observations_room)
             # self.assertIsNone(self.env.question)
@@ -458,7 +458,7 @@ def test_all(self) -> None:
         while True:
             actions_qa = [random.choice(observations["questions"][0])]
             action_explore = random.choice(["north", "east", "south", "west", "stay"])
-            observations, reward, done, truncated, info = self.env.step(
+            observations, rewards_, done, truncated, info = self.env.step(
                 (actions_qa, action_explore)
             )
             if done: