Import example learning agents

.github/workflows/python.yml

@@ -0,0 +1,67 @@
+################################################################################
+#
+#  Copyright (C) 2024 retro.ai
+#  This file is part of retro3 - https://github.com/retroai/retro3
+#
+#  SPDX-License-Identifier: AGPL-3.0-or-later
+#  See the file LICENSE for more information.
+#
+################################################################################
+
+name: Python CI
+
+on: [push, pull_request]
+
+jobs:
+  build-and-deploy:
+    # The type of runner that the job will run on
+    runs-on: ${{ matrix.os }}
+
+    defaults:
+      run:
+        # Set the working directory to frontend folder
+        working-directory: src/learning
+
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - os: ubuntu-latest
+            python-version: '3.11'
+
+    steps:
+    - name: Checkout 🛎️
+      uses: actions/checkout@v4
+
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v4
+      with:
+        python-version: ${{ matrix.python-version }}
+
+    - name: Install poetry
+      run: |
+        curl -sSL https://install.python-poetry.org | python3 -
+
+    - name: Configure Poetry
+      run: |
+        poetry config virtualenvs.create false
+
+    - name: Install dependencies
+      run: |
+        poetry install --no-root
+
+    - name: Check formatting with Black
+      run: |
+        poetry run black --check .
+
+    - name: Lint with Flake8
+      run: |
+        poetry run flake8 .
+
+    - name: Sort imports with isort
+      run: |
+        poetry run isort . --check-only
+
+    - name: Type check with mypy
+      run: |
+        poetry run mypy .

README.md

@@ -41,6 +41,16 @@ cmake .
 make -j$(nproc)
 ```
 
+## Testing
+
+Once `make` has completed in the `openai` folder, try running the two example learners:
+
+```bash
+cd src/learning
+./random_agent.py
+./brute_agent.py
+```
+
 ## Repo layout
 
 The following subfolders compose the repo architecture:

openai/retro/data/__init__.py

@@ -10,7 +10,7 @@
     from enum import Flag
 except ImportError:
     # Python < 3.6 doesn't support Flag, so we polyfill it ourself
-    class Flag(enum.Enum):
+    class Flag(enum.Enum):  # type: ignore
         def __and__(self, b):
             value = self.value & b.value
             try:

src/learning/.gitignore

@@ -0,0 +1 @@
+__pycache__

src/learning/brute_agent.py

@@ -0,0 +1,55 @@
+#!/usr/bin/env python3
+################################################################################
+#
+#  Copyright (C) 2023-2024 retro.ai
+#  This file is part of retro3 - https://github.com/retroai/retro3
+#
+#  This file is derived from OpenAI's Gym Retro under the MIT license
+#  Copyright (c) 2017-2018 OpenAI (http://openai.com)
+#
+#  SPDX-License-Identifier: AGPL-3.0-or-later AND MIT
+#  See the file LICENSE for more information.
+#
+################################################################################
+
+import retro.data  # noqa: F401
+import retroai.brute as brute_module
+import retroai.enums
+import retroai.retro_env
+
+
+def main():
+    game = "Airstriker-Genesis"
+    state = retroai.enums.State.DEFAULT
+    scenario = None
+    max_episode_steps = 4500
+    timestep_limit = 1e8
+
+    env = retroai.retro_env.retro_make(
+        game,
+        state,
+        use_restricted_actions=retroai.enums.Actions.DISCRETE,
+        scenario=scenario,
+    )
+
+    env = brute_module.Frameskip(env)
+    env = brute_module.TimeLimit(env, max_episode_steps=max_episode_steps)
+
+    brute = brute_module.Brute(env, max_episode_steps=max_episode_steps)
+    timesteps = 0
+    best_rew = float("-inf")
+    while True:
+        acts, rew = brute.run()
+        timesteps += len(acts)
+
+        if rew > best_rew:
+            print("new best reward {} => {}".format(best_rew, rew))
+            best_rew = rew
+
+        if timesteps > timestep_limit:
+            print("timestep limit exceeded")
+            break
+
+
+if __name__ == "__main__":
+    main()