Merge pull request #115 from FLAIROx/req-update

Update requirement structure, increment to new version.

Dockerfile

@@ -17,7 +17,7 @@ RUN apt-get update && \
     apt-get install -y tmux
 
 #jaxmarl from source if needed, all the requirements
-RUN pip install -e .
+RUN pip install -e .[algs,dev]
 
 USER ${MYUSER}
 

README.md

@@ -36,6 +36,8 @@
 
 ## Multi-Agent Reinforcement Learning in JAX
 
+🎉 **Update: JaxMARL was accepted at NeurIPS 2024 on Datasets and Benchmarks Track. See you in Vacouver!**
+
 JaxMARL combines ease-of-use with GPU-enabled efficiency, and supports a wide range of commonly used MARL environments as well as popular baseline algorithms. Our aim is for one library that enables thorough evaluation of MARL methods across a wide range of tasks and against relevant baselines. We also introduce SMAX, a vectorised, simplified version of the popular StarCraft Multi-Agent Challenge, which removes the need to run the StarCraft II game engine. 
 
 For more details, take a look at our [blog post](https://blog.foersterlab.com/jaxmarl/) or our [Colab notebook](https://colab.research.google.com/github/FLAIROx/JaxMARL/blob/main/jaxmarl/tutorials/JaxMARL_Walkthrough.ipynb), which walks through the basic usage.
@@ -72,7 +74,7 @@ We follow CleanRL's philosophy of providing single file implementations which ca
 
 <h2 name="install" id="install">Installation 🧗 </h2>
 
-**Environments** - Before installing, ensure you have the correct [JAX version](https://github.com/google/jax#installation) for your hardware accelerator. The JaxMARL environments can be installed directly from PyPi:
+**Environments** - Before installing, ensure you have the correct [JAX installation](https://github.com/google/jax#installation) for your hardware accelerator. We have tested up to JAX version 0.4.25. The JaxMARL environments can be installed directly from PyPi:
 
 ```
 pip install jaxmarl 
@@ -84,11 +86,15 @@ pip install jaxmarl
     ```
     git clone https://github.com/FLAIROx/JaxMARL.git && cd JaxMARL
     ```
-2. The requirements for IPPO & MAPPO can be installed with:
+2. Install requirements:
     ``` 
-    pip install -e .
+    pip install -e .[algs]
     export PYTHONPATH=./JaxMARL:$PYTHONPATH
     ```
+3. For the fastest start, we reccoment using our Dockerfile, the usage of which is outlined below.
+
+**Development** - If you would like to run our test suite, install the additonal dependencies with:
+ `pip install -e .[dev]`, after cloning the repository.
 
 <h2 name="start" id="start">Quick Start 🚀 </h2>
 
@@ -151,10 +157,12 @@ JAX-native algorithms:
 - [Mava](https://github.com/instadeepai/Mava): JAX implementations of IPPO and MAPPO, two popular MARL algorithms.
 - [PureJaxRL](https://github.com/luchris429/purejaxrl): JAX implementation of PPO, and demonstration of end-to-end JAX-based RL training.
 - [Minimax](https://github.com/facebookresearch/minimax/): JAX implementations of autocurricula baselines for RL.
+- [JaxIRL](https://github.com/FLAIROx/jaxirl?tab=readme-ov-file): JAX implementation of algorithms for inverse reinforcement learning.
 
 JAX-native environments:
 - [Gymnax](https://github.com/RobertTLange/gymnax): Implementations of classic RL tasks including classic control, bsuite and MinAtar.
 - [Jumanji](https://github.com/instadeepai/jumanji): A diverse set of environments ranging from simple games to NP-hard combinatorial problems.
 - [Pgx](https://github.com/sotetsuk/pgx): JAX implementations of classic board games, such as Chess, Go and Shogi.
 - [Brax](https://github.com/google/brax): A fully differentiable physics engine written in JAX, features continuous control tasks.
 - [XLand-MiniGrid](https://github.com/corl-team/xland-minigrid): Meta-RL gridworld environments inspired by XLand and MiniGrid.
+- [Craftax](https://github.com/MichaelTMatthews/Craftax): (Crafter + NetHack) in JAX.

jaxmarl/__init__.py

@@ -1,4 +1,4 @@
 from .registration import make, registered_envs
 
 __all__ = ["make", "registered_envs"]
-__version__ = "0.0.5"
+__version__ = "0.0.6"

jaxmarl/environments/hanabi/hanabi.py

@@ -9,7 +9,7 @@
 import chex
 from typing import Tuple, Dict
 from functools import partial
-from gymnax.environments.spaces import Discrete
+from jaxmarl.environments.spaces import Discrete
 from .hanabi_game import HanabiGame, State
 
 

jaxmarl/environments/mabrax/mabrax_env.py

@@ -1,7 +1,7 @@
 from typing import Dict, Literal, Optional, Tuple
 import chex
 from jaxmarl.environments.multi_agent_env import MultiAgentEnv
-from gymnax.environments import spaces
+from jaxmarl.environments import spaces
 from brax import envs
 import jax
 import jax.numpy as jnp

jaxmarl/environments/mpe/simple.py

@@ -10,7 +10,7 @@
 from jaxmarl.environments.multi_agent_env import MultiAgentEnv
 from jaxmarl.environments.mpe.default_params import *
 import chex
-from gymnax.environments.spaces import Box, Discrete
+from jaxmarl.environments.spaces import Box, Discrete
 from flax import struct
 from typing import Tuple, Optional, Dict
 from functools import partial

jaxmarl/environments/mpe/simple_adversary.py

@@ -5,7 +5,7 @@
 from functools import partial
 from jaxmarl.environments.mpe.simple import State, SimpleMPE
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box
+from jaxmarl.environments.spaces import Box
 
 
 class SimpleAdversaryMPE(SimpleMPE):

jaxmarl/environments/mpe/simple_crypto.py

@@ -6,7 +6,7 @@
 from functools import partial
 from jaxmarl.environments.mpe.simple import SimpleMPE, State
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box, Discrete
+from jaxmarl.environments.spaces import Box, Discrete
 
 SPEAKER = "alice_0"
 LISTENER = "bob_0"

jaxmarl/environments/mpe/simple_facmac.py

@@ -4,7 +4,7 @@
 from typing import Tuple, Dict
 from functools import partial
 from jaxmarl.environments.mpe.simple import State, SimpleMPE
-from gymnax.environments.spaces import Box
+from jaxmarl.environments.spaces import Box
 from jaxmarl.environments.mpe.default_params import *
 
 

jaxmarl/environments/mpe/simple_push.py

@@ -5,7 +5,7 @@
 from functools import partial
 from jaxmarl.environments.mpe.simple import SimpleMPE, State
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box
+from jaxmarl.environments.spaces import Box
 
 # Obstacle Colours
 COLOUR_1 = jnp.array([0.1, 0.9, 0.1])

jaxmarl/environments/mpe/simple_reference.py

@@ -5,7 +5,7 @@
 from functools import partial
 from jaxmarl.environments.mpe.simple import SimpleMPE, State
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box, Discrete
+from jaxmarl.environments.spaces import Box, Discrete
 
 # Obstacle Colours
 OBS_COLOUR = [(191, 64, 64), (64, 191, 64), (64, 64, 191)]

jaxmarl/environments/mpe/simple_speaker_listener.py

@@ -4,7 +4,7 @@
 from typing import Tuple, Dict
 from jaxmarl.environments.mpe.simple import SimpleMPE, State
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box, Discrete
+from jaxmarl.environments.spaces import Box, Discrete
 
 SPEAKER = "speaker_0"
 LISTENER = "listener_0"

jaxmarl/environments/mpe/simple_spread.py

@@ -5,7 +5,7 @@
 from functools import partial
 from jaxmarl.environments.mpe.simple import SimpleMPE, State
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box
+from jaxmarl.environments.spaces import Box
 
 
 class SimpleSpreadMPE(SimpleMPE):

jaxmarl/environments/mpe/simple_tag.py

@@ -4,7 +4,7 @@
 from typing import Tuple, Dict
 from functools import partial
 from jaxmarl.environments.mpe.simple import SimpleMPE, State
-from gymnax.environments.spaces import Box
+from jaxmarl.environments.spaces import Box
 from jaxmarl.environments.mpe.default_params import *
 
 

jaxmarl/environments/mpe/simple_world_comm.py

@@ -11,8 +11,7 @@
     OBS_COLOUR,
 )
 from jaxmarl.environments.mpe.default_params import *
-from gymnax.environments.spaces import Box, Discrete
-
+from jaxmarl.environments.spaces import Box, Discrete
 
 # NOTE food and forests are part of world.landmarks
 

jaxmarl/environments/spaces.py

@@ -1,3 +1,4 @@
+""" Built off Gymnax spaces.py, this module contains jittable classes for action and observation spaces. """
 from typing import Tuple, Union, Sequence
 from collections import OrderedDict
 import chex

pyproject.toml

@@ -8,7 +8,6 @@ include = ['jaxmarl*']
 
 [tool.setuptools.dynamic]
 version = {attr = "jaxmarl.__version__"}
-dependencies = {file = ["requirements/requirements.txt"]}
 
 [project]
 name = "jaxmarl"
@@ -17,7 +16,7 @@ description = "Multi-Agent Reinforcement Learning with JAX"
 authors = [
     {name = "Foerster Lab for AI Research", email = "arutherford@robots.ox.ac.uk"},
   ]
-dynamic = ["version", "dependencies"]
+dynamic = ["version"]
 license = {file = "LICENSE"}
 requires-python = ">=3.10"
 classifiers = [
@@ -31,6 +30,35 @@ classifiers = [
     "Topic :: Software Development :: Libraries :: Python Modules",
     "License :: OSI Approved :: Apache Software License",
 ]
+dependencies = [
+  "jax>=0.4.16.0,<=0.4.25",
+  "jaxlib>=0.4.16.0,<=0.4.25",
+  "flax",
+  "safetensors",
+  "chex",
+  "brax==0.10.3",
+  "mujoco==3.1.3",
+  "matplotlib",
+  "pillow",
+  "scipy<=1.12",
+  "gymnax",
+]
+
+[project.optional-dependencies]
+algs = [
+  "optax",
+  "distrax",
+  "flashbax==0.1.0",
+  "wandb",
+  "hydra-core>=1.3.2",
+  "omegaconf>=2.3.0",
+  "pettingzoo>=1.24.3",
+  "tqdm>=4.66.0",
+]
+dev = [
+  "pytest",
+  "pygame",
+]
 
 [project.urls]
 "Homepage" = "https://github.com/FLAIROx/JaxMARL"

tests/hanabi/test_hanabi.py

@@ -4,7 +4,6 @@
 import jax
 from jax import numpy as jnp
 from jaxmarl import make
-from jaxmarl.wrappers.baselines import LogWrapper
 
 env = make("hanabi")
 dir_path = os.path.dirname(os.path.realpath(__file__))