Finish refactoring to match src layout.

MatthewGerber · Jul 6, 2024 · e543c1b · e543c1b
1 parent e69d2a4
commit e543c1b
Show file tree

Hide file tree

Showing 10 changed files with 30 additions and 33 deletions.
diff --git a/...rlai/core/environments/openai_gym_test.py → .../rlai/core/environments/gymnasium_test.py b/...rlai/core/environments/openai_gym_test.py → .../rlai/core/environments/gymnasium_test.py
diff --git a/test/rlai/rewards/reward_test.py → test/rlai/core/reward_test.py b/test/rlai/rewards/reward_test.py → test/rlai/core/reward_test.py
diff --git a/test/rlai/states/test_states.py → test/rlai/core/state_test.py b/test/rlai/states/test_states.py → test/rlai/core/state_test.py
diff --git a/.../test_resume_gym_valid_environment.pickle → ...xtures/test_resume_from_checkpoint.pickle b/.../test_resume_gym_valid_environment.pickle → ...xtures/test_resume_from_checkpoint.pickle
diff --git a/test/rlai/gpi/fixtures/test_resume_gym_invalid_environment.pickle b/test/rlai/gpi/fixtures/test_resume_gym_invalid_environment.pickle
diff --git a/test/rlai/gpi/improvement_test.py b/test/rlai/gpi/improvement_test.py
diff --git a/test/rlai/gpi/state_action_value/tabular_test.py b/test/rlai/gpi/state_action_value/tabular_test.py
@@ -1,6 +1,9 @@
 import numpy as np
 import pytest
+from numpy.random import RandomState
 
+from rlai.core.environments.gridworld import Gridworld
+from rlai.gpi.state_action_value import ActionValueMdpAgent
 from rlai.gpi.state_action_value.tabular import TabularPolicy
 from rlai.gpi.state_action_value.tabular import TabularStateActionValueEstimator
 
@@ -39,3 +42,27 @@ def test_invalid_epsilon():
 
     with pytest.raises(ValueError, match='epsilon must be >= 0'):
         TabularStateActionValueEstimator(None, -1, None)
+
+
+def test_invalid_improve_policy_with_q_pi():
+    """
+    Test.
+    """
+
+    random_state = RandomState(12345)
+    mdp_environment: Gridworld = Gridworld.example_4_1(random_state, None)
+    epsilon = 0.0
+    mdp_agent = ActionValueMdpAgent(
+        'test',
+        random_state,
+        1,
+        TabularStateActionValueEstimator(mdp_environment, epsilon, None)
+    )
+
+    assert isinstance(mdp_agent.pi, TabularPolicy)
+
+    with pytest.raises(ValueError, match='Epsilon must be >= 0'):
+        mdp_agent.pi.improve_with_q_pi(
+            {},
+            -1.0
+        )
diff --git a/test/rlai/gpi/utils_test.py b/test/rlai/gpi/utils_test.py
@@ -13,7 +13,7 @@
 from test.rlai.utils import start_virtual_display_if_headless
 
 
-def test_resume_gym_valid_environment():
+def test_resume_from_checkpoint():
     """
     Test.
     """
@@ -49,10 +49,10 @@ def train_function_args_callback(
     resume_environment.close()
 
     # uncomment the following line and run test to update fixture
-    # with open(f'{os.path.dirname(__file__)}/fixtures/test_resume_gym_valid_environment.pickle', 'wb') as file:
+    # with open(f'{os.path.dirname(__file__)}/fixtures/test_resume_from_checkpoint.pickle', 'wb') as file:
     #     pickle.dump(agent.pi, file)
 
-    with open(f'{os.path.dirname(__file__)}/fixtures/test_resume_gym_valid_environment.pickle', 'rb') as file:
+    with open(f'{os.path.dirname(__file__)}/fixtures/test_resume_from_checkpoint.pickle', 'rb') as file:
         pi_fixture = pickle.load(file)
 
     assert agent.pi == pi_fixture
diff --git a/test/rlai/rewards/__init__.py b/test/rlai/rewards/__init__.py
diff --git a/test/rlai/states/__init__.py b/test/rlai/states/__init__.py