Skip to main content

AgileRL is a deep reinforcement learning library focused on improving RL development through RLOps.

Project description


Reinforcement learning streamlined.
Easier and faster reinforcement learning with RLOps. Visit our website. View documentation.
Join the Discord Server to collaborate.

License Documentation Status Downloads Discord

NEW: AgileRL now introduces evolvable Contextual Multi-armed Bandit Algorithms!

This is a Deep Reinforcement Learning library focused on improving development by introducing RLOps - MLOps for reinforcement learning.

This library is initially focused on reducing the time taken for training models and hyperparameter optimization (HPO) by pioneering evolutionary HPO techniques for reinforcement learning.
Evolutionary HPO has been shown to drastically reduce overall training times by automatically converging on optimal hyperparameters, without requiring numerous training runs.
We are constantly adding more algorithms and features. AgileRL already includes state-of-the-art evolvable on-policy, off-policy, offline, multi-agent and contextual multi-armed bandit reinforcement learning algorithms with distributed training.

AgileRL offers 10x faster hyperparameter optimization than SOTA.
Global steps is the sum of every step taken by any agent in the environment, including across an entire population, during the entire hyperparameter optimization process.

Table of Contents


Reinforcement learning algorithms and libraries are usually benchmarked once the optimal hyperparameters for training are known, but it often takes hundreds or thousands of experiments to discover these. This is unrealistic and does not reflect the true, total time taken for training. What if we could remove the need to conduct all these prior experiments?

In the charts below, a single AgileRL run, which automatically tunes hyperparameters, is benchmarked against Optuna's multiple training runs traditionally required for hyperparameter optimization, demonstrating the real time savings possible. Global steps is the sum of every step taken by any agent in the environment, including across an entire population.

AgileRL offers an order of magnitude speed up in hyperparameter optimization vs popular reinforcement learning training frameworks combined with Optuna. Remove the need for multiple training runs and save yourself hours.

AgileRL also supports multi-agent reinforcement learning using the Petting Zoo-style (parallel API). The charts below highlight the performance of our MADDPG and MATD3 algorithms with evolutionary hyper-parameter optimisation (HPO), benchmarked against epymarl's MADDPG algorithm with grid-search HPO for the simple speaker listener and simple spread environments.

Get Started

Install as a package with pip:

pip install agilerl

Or install in development mode:

git clone && cd AgileRL
pip install -e .


cd demos

or to demo distributed training:

cd demos
accelerate launch --config_file configs/accelerate/accelerate.yaml demos/


We are in the process of creating tutorials on how to use AgileRL and train agents on a variety of tasks.

Currently, we have tutorials for single-agent tasks that will guide you through the process of training both on and off-policy agents to beat a variety of Gymnasium environments. Additionally, we have multi-agent tutorials that make use of PettingZoo environments such as training DQN to play Connect Four with curriculum learning and self-play, and also for multi-agent tasks in MPE environments. We also have a tutorial on using hierarchical curriculum learning to teach agents Skills. We also have files for a tutorial on training a language model with reinforcement learning using ILQL on Wordle in tutorials/Language. If using ILQL on Wordle, download and unzip here.

Our demo files in demos also provide examples on how to train agents using AgileRL, and more information can be found in our documentation.

Evolvable algorithms implemented (more coming soon!)

  • DQN
  • Rainbow DQN
  • DDPG
  • TD3
  • PPO
  • CQL
  • ILQL
  • MATD3
  • NeuralUCB
  • NeuralTS

Train an agent to beat a Gym environment

Before starting training, there are some meta-hyperparameters and settings that must be set. These are defined in INIT_HP, for general parameters, and MUTATION_PARAMS, which define the evolutionary probabilities, and NET_CONFIG, which defines the network architecture. For example:

    'ENV_NAME': 'LunarLander-v2',   # Gym environment name
    'ALGO': 'DQN',                  # Algorithm
    'DOUBLE': True,                 # Use double Q-learning
    'CHANNELS_LAST': False,         # Swap image channels dimension from last to first [H, W, C] -> [C, H, W]
    'BATCH_SIZE': 256,              # Batch size
    'LR': 1e-3,                     # Learning rate
    'EPISODES': 2000,               # Max no. episodes
    'TARGET_SCORE': 200.,           # Early training stop at avg score of last 100 episodes
    'GAMMA': 0.99,                  # Discount factor
    'MEMORY_SIZE': 10000,           # Max memory buffer size
    'LEARN_STEP': 1,                # Learning frequency
    'TAU': 1e-3,                    # For soft update of target parameters
    'TOURN_SIZE': 2,                # Tournament size
    'ELITISM': True,                # Elitism in tournament selection
    'POP_SIZE': 6,                  # Population size
    'EVO_EPOCHS': 20,               # Evolution frequency
    'POLICY_FREQ': 2,               # Policy network update frequency
    'WANDB': True                   # Log with Weights and Biases
    # Relative probabilities
    'NO_MUT': 0.4,                              # No mutation
    'ARCH_MUT': 0.2,                            # Architecture mutation
    'NEW_LAYER': 0.2,                           # New layer mutation
    'PARAMS_MUT': 0.2,                          # Network parameters mutation
    'ACT_MUT': 0,                               # Activation layer mutation
    'RL_HP_MUT': 0.2,                           # Learning HP mutation
    'RL_HP_SELECTION': ['lr', 'batch_size'],    # Learning HPs to choose from
    'MUT_SD': 0.1,                              # Mutation strength
    'RAND_SEED': 1,                             # Random seed
    'arch': 'mlp',      # Network architecture
    'h_size': [32, 32], # Actor hidden size

First, use utils.utils.initialPopulation to create a list of agents - our population that will evolve and mutate to the optimal hyperparameters.

from agilerl.utils.utils import makeVectEnvs, initialPopulation
import torch

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

env = makeVectEnvs(env_name=INIT_HP['ENV_NAME'], num_envs=16)
    state_dim = env.single_observation_space.n          # Discrete observation space
    one_hot = True                                      # Requires one-hot encoding
except Exception:
    state_dim = env.single_observation_space.shape      # Continuous observation space
    one_hot = False                                     # Does not require one-hot encoding
    action_dim = env.single_action_space.n             # Discrete action space
except Exception:
    action_dim = env.single_action_space.shape[0]      # Continuous action space

    state_dim = (state_dim[2], state_dim[0], state_dim[1])

agent_pop = initialPopulation(algo=INIT_HP['ALGO'],                 # Algorithm
                              state_dim=state_dim,                  # State dimension
                              action_dim=action_dim,                # Action dimension
                              one_hot=one_hot,                      # One-hot encoding
                              net_config=NET_CONFIG,                # Network configuration
                              INIT_HP=INIT_HP,                      # Initial hyperparameters
                              population_size=INIT_HP['POP_SIZE'],  # Population size

Next, create the tournament, mutations and experience replay buffer objects that allow agents to share memory and efficiently perform evolutionary HPO.

from agilerl.components.replay_buffer import ReplayBuffer
from agilerl.hpo.tournament import TournamentSelection
from agilerl.hpo.mutation import Mutations

field_names = ["state", "action", "reward", "next_state", "done"]
memory = ReplayBuffer(action_dim=action_dim,                # Number of agent actions
                      memory_size=INIT_HP['MEMORY_SIZE'],   # Max replay buffer size
                      field_names=field_names,              # Field names to store in memory

tournament = TournamentSelection(tournament_size=INIT_HP['TOURN_SIZE'], # Tournament selection size
                                 elitism=INIT_HP['ELITISM'],            # Elitism in tournament selection
                                 population_size=INIT_HP['POP_SIZE'],   # Population size
                                 evo_step=INIT_HP['EVO_EPOCHS'])        # Evaluate using last N fitness scores

mutations = Mutations(algo=INIT_HP['ALGO'],                                 # Algorithm
                      no_mutation=MUTATION_PARAMS['NO_MUT'],                # No mutation
                      architecture=MUTATION_PARAMS['ARCH_MUT'],             # Architecture mutation
                      new_layer_prob=MUTATION_PARAMS['NEW_LAYER'],          # New layer mutation
                      parameters=MUTATION_PARAMS['PARAMS_MUT'],             # Network parameters mutation
                      activation=MUTATION_PARAMS['ACT_MUT'],                # Activation layer mutation
                      rl_hp=MUTATION_PARAMS['RL_HP_MUT'],                   # Learning HP mutation
                      rl_hp_selection=MUTATION_PARAMS['RL_HP_SELECTION'],   # Learning HPs to choose from
                      mutation_sd=MUTATION_PARAMS['MUT_SD'],                # Mutation strength
                      arch=NET_CONFIG['arch'],                              # Network architecture
                      rand_seed=MUTATION_PARAMS['RAND_SEED'],               # Random seed

The easiest training loop implementation is to use our train_off_policy() function. It requires the agent have functions getAction() and learn().

from import train_off_policy

trained_pop, pop_fitnesses = train_off_policy(env=env,                                 # Gym-style environment
                                   env_name=INIT_HP['ENV_NAME'],            # Environment name
                                   algo=INIT_HP['ALGO'],                    # Algorithm
                                   pop=agent_pop,                           # Population of agents
                                   memory=memory,                           # Replay buffer
                                   swap_channels=INIT_HP['CHANNELS_LAST'],  # Swap image channel from last to first
                                   n_episodes=INIT_HP['EPISODES'],          # Max number of training episodes
                                   evo_epochs=INIT_HP['EVO_EPOCHS'],        # Evolution frequency
                                   evo_loop=1,                              # Number of evaluation episodes per agent
                                   target=INIT_HP['TARGET_SCORE'],          # Target score for early stopping
                                   tournament=tournament,                   # Tournament selection object
                                   mutation=mutations,                      # Mutations object
                                   wb=INIT_HP['WANDB'])                     # Weights and Biases tracking

Citing AgileRL

If you use AgileRL in your work, please cite the repository:

author = {Ustaran-Anderegg, Nicholas and Pratt, Michael},
license = {Apache-2.0},
title = {{AgileRL}},
url = {}

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

agilerl-0.1.21.tar.gz (146.1 kB view hashes)

Uploaded source

Built Distribution

agilerl-0.1.21-py3-none-any.whl (189.3 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page