Merge pull request #53 from bmielnicki/python_state_visualization

Python native state visualization

Author: micahcarroll

.github/workflows/pythontests.yml

@@ -12,11 +12,10 @@ jobs:
     - name: Install dependencies
       run: python -m pip install --upgrade pip
     - name: Setup
-      run: pip install .
+      run: pip install -e .
     - name: Run tests
       run: |
         python -m unittest discover -s testing/ -p "*_test.py"
-
   ubuntu_mdp_tests:
     runs-on: ubuntu-latest
     steps:
@@ -34,8 +33,7 @@ jobs:
       run: pip install -e .
     - name: Run tests and generate coverage report
       run: |
-        coverage run -m unittest discover -s testing/ -p "*_test.py"
-        coverage report
+        python -m unittest discover -s testing/ -p "*_test.py"
     - name: Upload coverage to Codecov
       uses: codecov/codecov-action@v1
       with:

setup.py

@@ -24,6 +24,8 @@
         'numpy',
         'tqdm',
         'gym',
-        'ipython'
+        'ipython',
+        'pygame',
+        "ipywidgets"
       ]
     )

src/overcooked_ai_py/agents/agent.py

@@ -458,7 +458,20 @@ def ml_action(self, state):
 
         return motion_goals
 
+class SampleAgent(Agent):
+    """ Agent that samples action using the average action_probs across multiple agents
+    """
+    def __init__(self, agents):
+        self.agents = agents
 
+    def action(self, state):
+        action_probs = np.zeros(Action.NUM_ACTIONS)
+        for agent in self.agents:
+            action_probs += agent.action(state)[1]["action_probs"]
+        action_probs = action_probs/len(self.agents)
+        return Action.sample(action_probs), {"action_probs": action_probs}
+    """
+    """
 # Deprecated. Need to fix Heuristic to work with the new MDP to reactivate Planning
 # class CoupledPlanningAgent(Agent):
 #     """