multi-agent-rlenv 3.6.1__tar.gz → 3.6.2__tar.gz

{multi_agent_rlenv-3.6.1 → multi_agent_rlenv-3.6.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: multi-agent-rlenv
-Version: 3.6.1
+Version: 3.6.2
 Summary: A strongly typed Multi-Agent Reinforcement Learning framework
 Project-URL: repository, https://github.com/yamoling/multi-agent-rlenv
 Author-email: Yannick Molinghen <yannick.molinghen@ulb.be>

{multi_agent_rlenv-3.6.1 → multi_agent_rlenv-3.6.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "multi-agent-rlenv"
-version = "3.6.1"
+version = "3.6.2"
 description = "A strongly typed Multi-Agent Reinforcement Learning framework"
 authors = [
     { "name" = "Yannick Molinghen", "email" = "yannick.molinghen@ulb.be" },

{multi_agent_rlenv-3.6.1 → multi_agent_rlenv-3.6.2}/src/marlenv/catalog/deepsea.py

@@ -34,7 +34,7 @@ class DeepSea(MARLEnv[MultiDiscreteSpace]):
         self._step_right_penalty = -0.01 / self.max_depth
 
     def get_observation(self) -> Observation:
-        return Observation(np.array([self._row, self._col], dtype=np.float32), self.available_actions())
+        return Observation(np.array([[self._row, self._col]], dtype=np.float32), self.available_actions())
 
     def get_state(self) -> State:
         return State(np.array([self._row, self._col], dtype=np.float32))

multi_agent_rlenv-3.6.2/tests/test_deepsea.py

@@ -0,0 +1,43 @@
+from marlenv.catalog.deepsea import DeepSea, LEFT, RIGHT
+
+
+def test_env():
+    env = DeepSea(20)
+    assert env.n_actions == 2
+    assert env.action_space.is_discrete
+    assert env.observation_shape == (2,)
+    assert env.n_agents == 1
+
+
+def test_reset():
+    env = DeepSea(20)
+    obs, state = env.reset()
+    assert obs.shape == (2,)
+    assert state.shape == (2,)
+
+    assert obs.data[0][0] == 0
+    assert state.data[0] == 0
+    assert obs.data[0][1] == 0
+    assert state.data[1] == 0
+
+
+def test_step():
+    env = DeepSea(20)
+    env.reset()
+    step = env.step([RIGHT])
+    obs = step.obs
+    assert obs.data[0][0] == 1
+    assert obs.data[0][1] == 1
+    assert step.reward.item() < 0
+
+    step = env.step([LEFT])
+    obs = step.obs
+    assert obs.data[0][0] == 2
+    assert obs.data[0][1] == 0
+    assert step.reward.item() == 0.0
+
+    step = env.step([LEFT])
+    obs = step.obs
+    assert obs.data[0][0] == 3
+    assert obs.data[0][1] == 0
+    assert step.reward.item() == 0.0