WIP migrate to uv

.github/workflows/blackisort.yml

@@ -8,15 +8,23 @@ jobs:
     steps:
       - uses: actions/checkout@v3
       - uses: actions/setup-python@v4
-      # https://black.readthedocs.io/en/stable/integrations/github_actions.html
-      - name: running Black
-        uses: psf/black@stable
         with:
-          options: "--check --diff --line-length 79"
-          version: "23.3.0"
-      # https://pycqa.github.io/isort/docs/configuration/github_action.html
-      - name: running isort
-        uses: jamescurtin/isort-action@master
+          python-version: 3.8
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+      - name: Setup virtual environment and install black and isort
+        run: |
+          uv venv
+          source .venv/bin/activate
+          uv pip install black==23.3.0 isort
+      - name: Running Black
+        run: |
+          source .venv/bin/activate
+          black --check --diff --line-length 79 .
+      - name: Running isort
         if: always()
-        with:
-          configuration: "--check-only --line-length 79 --profile black"
+        run: |
+          source .venv/bin/activate
+          isort --check-only --line-length 79 --profile black .

.github/workflows/pythonlint.yml

@@ -13,11 +13,17 @@ jobs:
       with:
         python-version: 3.8
         architecture: 'x64'
-    - name: Install dependencies
+    - name: Install uv
       run: |
-        python -m pip install --upgrade pip
-        pip install flake8
+        curl -LsSf https://astral.sh/uv/install.sh | sh
+        echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+    - name: Setup virtual environment and install dependencies
+      run: |
+        uv venv
+        source .venv/bin/activate
+        uv pip install flake8
     - name: Lint with flake8
       run: |
+        source .venv/bin/activate
         # stop the build if there are Python syntax errors or undefined names
         flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics

.github/workflows/pythontests.yml

@@ -26,14 +26,22 @@ jobs:
       with:
         python-version: 3.8
         architecture: 'x64'
-    - name: Install dependencies
+    - name: Install uv
       run: |
-        python -m pip install --upgrade pip
-        pip install coverage
+        curl -LsSf https://astral.sh/uv/install.sh | sh
+        echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+    - name: Setup virtual environment and install dependencies
+      run: |
+        uv venv
+        source .venv/bin/activate
+        uv pip install coverage
     - name: Setup # needs -e flag for coverage tests to work
-      run: pip install -e .[harl]
+      run: |
+        source .venv/bin/activate
+        uv pip install -e ".[harl]"
     - name: Run tests and generate coverage report
       run: |
+        source .venv/bin/activate
         coverage run -m unittest discover -s testing/ -p "*_test.py"
     - name: Upload coverage to Codecov
       uses: codecov/codecov-action@v1

.gitignore

@@ -172,4 +172,7 @@ node_modules/
 **/__pycache__/
 #manually correcting for demo 
 !**/static/**/*.png
-!src/overcooked_demo/server/static/lib
+!src/overcooked_demo/server/static/lib
+
+# Ignore copied overcooked_ai directory in server
+src/overcooked_demo/server/overcooked_ai/

.uvrc

@@ -0,0 +1,9 @@
+[application]
+default-python = "3.8"
+isolated = true
+
+[venv]
+location = ".venv"
+
+[pypi]
+url = "https://pypi.org/simple" 

UV_MIGRATION.md

@@ -0,0 +1,55 @@
+# Migrating to uv Package Manager
+
+This project has been configured to use [uv](https://github.com/astral-sh/uv), a fast Python package manager. The following guide explains how to set up and use uv with this project.
+
+## Installing uv
+
+To install uv, run:
+
+```sh
+curl -LsSf https://astral.sh/uv/install.sh | sh
+```
+
+This adds uv to your system. You may need to restart your shell or add it to your PATH.
+
+## Setup for Development
+
+To set up a development environment:
+
+```sh
+# Clone the repository
+git clone https://github.com/HumanCompatibleAI/overcooked_ai.git
+cd overcooked_ai
+
+# Create a virtual environment and install dependencies
+uv venv
+source .venv/bin/activate  # On Windows: .venv\Scripts\activate
+
+# Install the project in development mode with all dependencies
+uv pip install -e ".[harl]"
+```
+
+## Common uv Commands
+
+Replace pip commands with uv equivalents:
+
+| pip | uv |
+|-----|-----|
+| `pip install package` | `uv pip install package` |
+| `pip install -r requirements.txt` | `uv pip install -r requirements.txt` |
+| `pip freeze > requirements.txt` | `uv pip freeze > requirements.txt` |
+
+## Benefits of uv
+
+- Faster installation speed
+- Better dependency resolution
+- Improved caching
+- Single binary with no dependencies
+
+## Project Structure
+
+The project now uses a `pyproject.toml` file for configuration, which is compatible with uv and modern Python packaging standards. The `setup.py` file has been maintained for backward compatibility.
+
+## CI/CD Integration
+
+GitHub Actions workflows have been updated to use uv for dependency installation. 

create_agent.py

@@ -0,0 +1,21 @@
+#!/usr/bin/env python
+import pickle
+import os
+
+class RandomAI:
+    def action(self, state):
+        import numpy as np
+        return np.random.choice(['up', 'down', 'left', 'right', 'stay', 'interact'])
+
+    def reset(self):
+        pass
+
+if __name__ == "__main__":
+    # Create agent directory
+    os.makedirs('/app/data/agents/random_agent/agent', exist_ok=True)
+
+    # Save the agent
+    with open('/app/data/agents/random_agent/agent/agent.pickle', 'wb') as f:
+        pickle.dump(RandomAI(), f)
+
+    print("Agent created successfully!") 

pyproject.toml

@@ -0,0 +1,72 @@
+[build-system]
+requires = ["setuptools>=61.0"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "overcooked_ai"
+version = "1.1.0"
+description = "Cooperative multi-agent environment based on Overcooked"
+readme = "README.md"
+requires-python = ">=3.9"
+license = {file = "LICENSE"}
+authors = [
+    {name = "Micah Carroll", email = "mdc@berkeley.edu"}
+]
+urls = {homepage = "https://github.com/HumanCompatibleAI/overcooked_ai"}
+keywords = ["Overcooked", "AI", "Reinforcement Learning"]
+dependencies = [
+    "dill",
+    "numpy<2.0.0",
+    "scipy",
+    "tqdm",
+    "gymnasium",
+    "ipython",
+    "pygame",
+    "ipywidgets",
+    "opencv-python",
+    "eventlet>=0.39.0",
+    "flask",
+    "flask-socketio",
+]
+
+[project.optional-dependencies]
+harl = [
+    "wandb",
+    "GitPython",
+    "memory_profiler",
+    "sacred",
+    "pymongo",
+    "matplotlib",
+    "requests",
+    "seaborn==0.9.0",
+    "ray[rllib]>=2.5.0",
+    "protobuf",
+    "tensorflow>=2.14.0",
+]
+
+[project.scripts]
+overcooked-demo-up = "overcooked_demo:start_server"
+overcooked-demo-move = "overcooked_demo:move_agent"
+
+[tool.setuptools]
+package-dir = {"" = "src"}
+packages = {find = {where = ["src"]}}
+
+[tool.setuptools.package-data]
+"overcooked_ai_py" = [
+    "data/layouts/*.layout",
+    "data/planners/*.py",
+    "data/human_data/*.pickle",
+    "data/graphics/*.png",
+    "data/graphics/*.json",
+    "data/fonts/*.ttf",
+]
+"human_aware_rl" = [
+    "static/**/*.pickle",
+    "static/**/*.csv",
+    "ppo/trained_example/*.pkl",
+    "ppo/trained_example/*.json",
+    "ppo/trained_example/*/.is_checkpoint",
+    "ppo/trained_example/*/.tune_metadata",
+    "ppo/trained_example/*/checkpoint-500",
+] 

setup.py

@@ -39,14 +39,16 @@
     },
     install_requires=[
         "dill",
-        "numpy",
+        "numpy<2.0.0",
         "scipy",
         "tqdm",
         "gymnasium",
         "ipython",
         "pygame",
         "ipywidgets",
         "opencv-python",
+        "flask",
+        "flask-socketio",
     ],
     # removed overlapping dependencies
     extras_require={
@@ -59,9 +61,9 @@
             "matplotlib",
             "requests",
             "seaborn==0.9.0",
-            "ray[rllib]==2.0.0",
+            "ray[rllib]>=2.5.0",
             "protobuf",
-            "tensorflow==2.10",
+            "tensorflow>=2.14.0",
         ]
     },
     entry_points={

src/human_aware_rl/rllib/rllib.py

@@ -9,13 +9,16 @@
 import gymnasium
 import numpy as np
 import ray
-from ray.rllib.agents.ppo import PPOTrainer
+# Updated import for Ray 2.5+
+from ray.rllib.algorithms.ppo import PPO as PPOTrainer
 from ray.rllib.algorithms.callbacks import DefaultCallbacks
 from ray.rllib.env.multi_agent_env import MultiAgentEnv
 from ray.rllib.models import ModelCatalog
 from ray.tune.logger import UnifiedLogger
 from ray.tune.registry import register_env
-from ray.tune.result import DEFAULT_RESULTS_DIR
+
+# Define DEFAULT_RESULTS_DIR since it's no longer available in Ray 2.5+
+DEFAULT_RESULTS_DIR = os.path.join(os.path.expanduser("~"), "ray_results")
 
 from human_aware_rl.rllib.utils import (
     get_base_ae,

src/overcooked_demo/README.md

@@ -7,6 +7,9 @@ A web application where humans can play Overcooked with trained AI agents.
 
 * [Installation](#installation)
 * [Usage](#usage)
+  * [Docker Deployment](#docker-deployment)
+  * [Local Run with uv](#local-run-with-uv)
+  * [Command Line](#command-line)
 * [Dependencies](#dependencies)
 * [Using Pre-trained Agents](#using-pre-trained-agents)
 * [Updating](#updating)
@@ -17,8 +20,14 @@ A web application where humans can play Overcooked with trained AI agents.
 
 Building the server image requires [Docker](https://docs.docker.com/get-docker/)
 
+For local development with uv, you'll need:
+- Python 3.7 or later
+- [uv](https://github.com/astral-sh/uv) - Install with `pip install uv`
+
 ## Usage
 
+### Docker Deployment
+
 The server can be deployed locally using the driver script included in the repo. To run the production server, use the command
 ```bash
 ./up.sh production
@@ -36,6 +45,23 @@ In order to kill the production server, run
 ./down.sh
 ```
 
+### Local Run with uv
+
+To run the server locally using uv without Docker, use:
+
+```bash
+./up.sh local
+```
+
+This will:
+1. Create a virtual environment in the server directory if it doesn't exist
+2. Install all dependencies using uv
+3. Start the server on port 5000
+
+After running this command, navigate to http://localhost:5000
+
+### Command Line
+
 You can also start the server via the command line. After installing the `overcooked_ai` via pip, you can start the server by typing
 
 ```

src/overcooked_demo/docker-compose.yml

@@ -1,5 +1,3 @@
-version : '3.7'
-
 services:
     app:
         build: