AgileRL
<p align="center"> <img src=https://user-images.githubusercontent.com/47857277/222710068-e09a4e3c-368c-458a-9e01-b68674806887.png height="120"> </p> <p align="center"><b>Reinforcement learning streamlined.</b><br>Easier and faster reinforcement learning with RLOps. Visit our <a href="https://agilerl.com">website</a>. View <a href="https://docs.agilerl.com">documentation</a>.<br>Join the <a href="https://discord.gg/eB8HyTA2ux">Discord Server</a> for questions, help and collaboration.</p> <div align="center"> [](https://opensource.org/licenses/Apache-2.0) [](https://docs.agilerl.com/en/latest/?badge=latest) [](https://github.com/AgileRL/AgileRL/actions/workflows/linux-tests.yml) [](https://github.com/AgileRL/AgileRL/actions/workflows/macos-tests.yml) [](https://github.com/AgileRL/AgileRL/actions/workflows/windows-tests.yml) [](https://pypi.python.org/pypi/agilerl/) [](https://discord.gg/eB8HyTA2ux) [](https://arena.agilerl.com) <br> <h3><i>🚀 <b>Train super-fast for free on <a href="https://arena.agilerl.com">Arena</a>, the RLOps platform from AgileRL 🚀</b></i></h3> </div> <br> AgileRL 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](https://docs.agilerl.com/en/latest/evo_hyperparam_opt/index.html) for reinforcement learning.<br> Evolutionary HPO has been shown to drastically reduce overall training times by automatically converging on optimal hyperparameters, without requiring numerous training runs.<br> We are constantly adding more algorithms and features. AgileRL already includes state-of-the-art evolvable [on-policy](https://docs.agilerl.com/en/latest/on_policy/index.html), [off-policy](https://docs.agilerl.com/en/latest/off_policy/index.html), [offline](https://docs.agilerl.com/en/latest/offline_training/index.html), [multi-agent](https://docs.agilerl.com/en/latest/multi_agent_training/index.html) and [contextual multi-armed bandit](https://docs.agilerl.com/en/latest/bandits/index.html) reinforcement learning algorithms with [distributed training](https://docs.agilerl.com/en/latest/distributed_training/index.html). <p align="center"> <img src=https://user-images.githubusercontent.com/47857277/236407686-21363eb3-ffcf-419f-b019-0be4ddf1ed4a.gif width="100%" max-width="900"> </p> <p align="center">AgileRL offers 10x faster hyperparameter optimization than SOTA.</p> ## Table of Contents * [Get Started](#get-started) * [Benchmarks](#benchmarks) * [Tutorials](#tutorials) * [Algorithms implemented](#evolvable-algorithms-more-coming-soon) * [Train an agent](#train-an-agent-to-beat-a-gym-environment) * [Citing AgileRL](#citing-agilerl) ## Get Started To see the full AgileRL documentation, including tutorials, visit our [documentation site](https://docs.agilerl.com/). To ask questions and get help, collaborate, or discuss anything related to reinforcement learning, join the [AgileRL Discord Server](https://discord.gg/eB8HyTA2ux). Install as a package with pip: ```bash pip install agilerl ``` Or install in development mode: ```bash git clone https://github.com/AgileRL/AgileRL.git && cd AgileRL pip install -e . ``` If you wish to install all additional dependencies please specify `[all]` or if you want to install a specific family of dependencies specify that family directly. At present, we have just one family, `[llm]`, which contains the dependencies related to our LLM RFT algorithms (datasets, deepspeed, peft, transformers, vllm). ```bash pip install agilerl[all] ``` Or in development mode: ```bash pip install -e ".[all]" ``` To install the ``nightly`` version of AgileRL with the latest features, use: ```bash pip install git+https://github.com/AgileRL/AgileRL.git@nightly ``` ## Benchmarks 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. <p align="center"> <img src=https://user-images.githubusercontent.com/47857277/227481592-27a9688f-7c0a-4655-ab32-90d659a71c69.png min-width="100%" width="600"> </p> <p align="center">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.</p> 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. <p align="center"> <img src=https://github-production-user-asset-6210df.s3.amazonaws.com/118982716/264712154-4965ea5f-b777-423c-989b-e4db86eda3bd.png min-width="100%" width="700"> </p> ## Tutorials We are constantly updating our tutorials to showcase the latest features of AgileRL and how users can leverage our evolutionary HPO to achieve 10x faster hyperparameter optimization. Please see the available tutorials below. | Tutorial Type | Description | Tutorials | |---------------|-------------|-----------| | [Single-agent tasks](https://docs.agilerl.com/en/latest/tutorials/gymnasium/index.html) | Guides for training both on and off-policy agents to beat a variety of Gymnasium environments. | [PPO - Acrobot](https://docs.agilerl.com/en/latest/tutorials/gymnasium/agilerl_ppo_tutorial.html) <br> [TD3 - Lunar Lander](https://docs.agilerl.com/en/latest/tutorials/gymnasium/agilerl_td3_tutorial.html) <br> [Rainbow DQN - CartPole](https://docs.agilerl.com/en/latest/tutorials/gymnasium/agilerl_rainbow_dqn_tutorial.html) <br> [Recurrent PPO - Masked Pendulum](https://docs.agilerl.com/en/latest/tutorials/gymnasium/agilerl_recurrent_ppo_tutorial.html) | | [Multi-agent tasks](https://docs.agilerl.com/en/latest/tutorials/pettingzoo/index.html) | Use of PettingZoo environments such as training DQN to play Connect Four with curriculum learning and self-play, and for multi-agent tasks in MPE environments. | [DQN - Connect Four](https://docs.agilerl.com/en/latest/tutorials/pettingzoo/dqn.html) <br> [MADDPG - Space Invaders](https://docs.agilerl.com/en/latest/tutorials/pettingzoo/maddpg.html) <br> [MATD3 - Speaker Listener](https://docs.agilerl.com/en/latest/tutorials/pettingzoo/matd3.html) | | [Hierarchical curriculum learning](https://docs.agilerl.com/en/latest/tutorials/skills/index.html) | Shows how to teach agents Skills and combine them to achieve an end goal. | [PPO - Lunar Lander](https://docs.agilerl.com/en/latest/tutorials/skills/index.html) | | [Contextual multi-arm bandits](https://docs.agilerl.com/en/latest/tutorials/bandits/index.html) | Learn to make the correct decision in environments that only have one timestep. | [NeuralUCB - Iris Dataset](https://docs.agilerl.com/en/latest/tutorials/bandits/agilerl_neural_ucb_tutorial.html) <br> [NeuralTS - PenDigits](https://docs.agilerl.com/en/latest/tutorials/bandits/agilerl_neural_ts_tutorial.html) | | [Custom Modules & Networks](https://docs.agilerl.com/en/latest/tutorials/custom_networks/index.html) | Learn how to create custom evolvable modules and networks for RL algorithms. | [Dueling Distributional Q Network](https://docs.agilerl.com/en/latest/tutorials/custom_networks/agilerl_rainbow_tutorial.html) <br> [EvolvableSimBa](https://docs.agilerl.com/en/latest/tutorials/custom_networks/agilerl_simba_tutorial.html) | | [LLM Finetuning](https://docs.agilerl.com/en/latest/tutorials/llm_finetuning/index.html) | Learn how to finetune an LLM using AgileRL. | [GRPO](https://docs.agilerl.com/en/latest/tutorials/llm_finetuning/index.html) | ## Evolvable algorithms (more coming soon!) ### Single-agent algorithms | RL | Algorithm | | ---------- | --------- | | [On-Policy](https://docs.agilerl.com/en/latest/on_policy/index.html) | [Proximal Policy Optimization (PPO)](https://docs.agilerl.com/en/latest/api/algorithms/ppo.html) | | [Off-Policy](https://docs.agilerl.com/en/latest/off_policy/index.html) | [Deep Q Learning (DQN)](https://docs.agilerl.com/en/latest/api/algorithms/dqn.html) <br> [Rainbow DQN](https://docs.agilerl.com/en/latest/api/algorithms/dqn_rainbow.html) <br> [Deep Deterministic Policy Gradient (DDPG)](https://docs.agilerl.com/en/latest/api/algorithms/ddpg.html) <br> [Twin Delayed Deep Deterministic Policy Gradient (TD3)](https://docs.agilerl.com/en/latest/api/algorithms/td3.html) | | [Offline](https://docs.agilerl.com/en/latest/offline_training/index.html) | [Conservative Q-Learning (CQL)](https://docs.agilerl.com/en/latest/api/algorithms/cql.html) <br> [Implicit Language Q-Learning (ILQL)](https://docs.agilerl.com/en/latest/api/algorithms/ilql.html) | ### Multi-agent algorithms | RL | Algorithm | | ---------- | --------- | | [Multi-agent](https://docs.agilerl.com/en/latest/multi_agent_training/index.html) | [Multi-Agent Deep Deterministic Policy Gradient (MADDPG)](https://docs.agilerl.com/en/latest/api/algorithms/maddpg.html) <br> [Multi-Agent Twin-Delayed Deep Deterministic Policy Gradient (MATD3)](https://docs.agilerl.com/en/latest/api/algorithms/matd3.html) <br> [Independent Proximal Policy Optimization (IPPO)](https://docs.agilerl.com/en/latest/api/algorithms/ippo.html)| ### Contextual multi-armed bandit algorithms | RL | Algorithm | | ---------- | --------- | | [Bandits](https://docs.agilerl.com/en/latest/bandits/index.html) | [Neural Contextual Bandits with UCB-based Exploration (NeuralUCB)](https://docs.agilerl.com/en/latest/api/algorithms/neural_ucb.html) <br> [Neural Contextual Bandits with Thompson Sampling (NeuralTS)](https://docs.agilerl.com/en/latest/api/algorithms/neural_ts.html) | ### LLM Fine-tuning Algorithms | RL | Algorithm | | ---------- | --------- | | [On-Policy](https://docs.agilerl.com/en/latest/llm_finetuning/index.html) | [Group Relative Policy Optimization (GRPO)](https://docs.agilerl.com/en/latest/api/algorithms/grpo.html) | [Off-Policy](https://docs.agilerl.com/en/latest/llm_finetuning/index.html) | [Direct Preference Optimization (DPO)](https://docs.agilerl.com/en/latest/api/algorithms/dpo.html) ## 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 <code>INIT_HP</code>, for general parameters, and <code>MUTATION_PARAMS</code>, which define the evolutionary probabilities, and <code>NET_CONFIG</code>, which defines the network architecture. For example: <details> <summary>Basic Hyperparameters</summary> ```python INIT_HP = { 'ENV_NAME': 'LunarLander-v3', # 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 'MAX_STEPS': 1_000_000, # Max no. steps '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_STEPS': 10_000, # Evolution frequency 'EVAL_STEPS': None, # Evaluation steps 'EVAL_LOOP': 1, # Evaluation episodes 'LEARNING_DELAY': 1000, # Steps before starting learning 'WANDB': True, # Log with Weights and Biases } ``` </details> <details> <summary>Mutation Hyperparameters</summary> ```python MUTATION_PARAMS = { # 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 'MUT_SD': 0.1, # Mutation strength 'RAND_SEED': 1, # Random seed } ``` </details> <details> <summary>Basic Network Configuration</summary> ```python NET_CONFIG = { 'latent_dim': 16 'encoder_config': { 'hidden_size': [32] # Observation encoder configuration } 'head_config': { 'hidden_size': [32] # Network head configuration } } ``` </details> ### Creating a Population of Agents First, use <code>utils.utils.create_population</code> to create a list of agents - our population that will evolve and mutate to the optimal hyperparameters. <details> <summary>Population Creation Example</summary> ```python import torch from agilerl.utils.utils import ( make_vect_envs, create_population, observation_space_channels_to_first ) device = torch.device("cuda" if torch.cuda.is_available() else "cpu") num_envs = 16 env = make_vect_envs(env_name=INIT_HP['ENV_NAME'], num_envs=num_envs) observation_space = env.single_observation_space action_space = env.single_action_space if INIT_HP['CHANNELS_LAST']: observation_space = observation_space_channels_to_first(observation_space) agent_pop = create_population( algo=INIT_HP['ALGO'], # Algorithm observation_space=observation_space, # Observation space action_space=action_space, # Action space net_config=NET_CONFIG, # Network configuration INIT_HP=INIT_HP, # Initial hyperparameters population_size=INIT_HP['POP_SIZE'], # Population size num_envs=num_envs, # Number of vectorized environments device=device ) ``` </details> ### Initializing Evolutionary HPO Next, create the tournament, mutations and experience replay buffer objects that allow agents to share memory and efficiently perform evolutionary HPO. <details> <summary>Mutations and Tournament Selection Example</summary> ```python from agilerl.components.replay_buffer import ReplayBuffer from agilerl.hpo.tournament import TournamentSelection from agilerl.hpo.mutation import Mutations memory = ReplayBuffer( max_size=INIT_HP['MEMORY_SIZE'], # Max replay buffer size device=device, ) 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 eval_loop=INIT_HP['EVAL_LOOP'], # Evaluate using last N fitness scores ) mutations = Mutations( 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 mutation_sd=MUTATION_PARAMS['MUT_SD'], # Mutation strength rand_seed=MUTATION_PARAMS['RAND_SEED'], # Random seed device=device, ) ``` </details> ### Train A Population of Agents The easiest training loop implementation is to use our <code>train_off_policy()</code> function. It requires the <code>agent</code> have methods <code>get_action()</code> and <code>learn().</code> ```python from agilerl.training.train_off_policy 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 max_steps=INIT_HP["MAX_STEPS"], # Max number of training steps evo_steps=INIT_HP['EVO_STEPS'], # Evolution frequency eval_steps=INIT_HP["EVAL_STEPS"], # Number of steps in evaluation episode eval_loop=INIT_HP["EVAL_LOOP"], # Number of evaluation episodes learning_delay=INIT_HP['LEARNING_DELAY'], # Steps before starting learning 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: ```bibtex @software{Ustaran-Anderegg_AgileRL, author = {Ustaran-Anderegg, Nicholas and Pratt, Michael and Sabal-Bermudez, Jaime}, license = {Apache-2.0}, title = {{AgileRL}}, url = {https://github.com/AgileRL/AgileRL} } ```