Ape-X variations of DQN, DDPG, and QMIX (`APEX_DQN <https://github.com/ray-project/ray/blob/master/rllib/agents/dqn/apex.py>`__, `APEX_DDPG <https://github.com/ray-project/ray/blob/master/rllib/agents/ddpg/apex.py>`__, `APEX_QMIX <https://github.com/ray-project/ray/blob/master/rllib/agents/qmix/apex.py>`__) use a single GPU learner and many CPU workers for experience collection. Experience collection can scale to hundreds of CPU workers due to the distributed prioritization of experience prior to storage in replay buffers.
In IMPALA, a central learner runs SGD in a tight loop while asynchronously pulling sample batches from many actor processes. RLlib's IMPALA implementation uses DeepMind's reference `V-trace code <https://github.com/deepmind/scalable_agent/blob/master/vtrace.py>`__. Note that we do not provide a deep residual network out of the box, but one can be plugged in as a `custom model <rllib-models.html#custom-models-tensorflow>`__. Multiple learner GPUs and experience replay are also supported.
We include an asynchronous variant of Proximal Policy Optimization (PPO) based on the IMPALA architecture. This is similar to IMPALA but using a surrogate policy loss with clipping. Compared to synchronous PPO, APPO is more efficient in wall-clock time due to its use of asynchronous sampling. Using a clipped loss also allows for multiple SGD passes, and therefore the potential for better sample efficiency compared to IMPALA. V-trace can also be enabled to correct for off-policy samples.
APPO is not always more efficient; it is often better to simply use `PPO <rllib-algorithms.html#proximal-policy-optimization-ppo>`__ or `IMPALA <rllib-algorithms.html#importance-weighted-actor-learner-architecture-impala>`__.
RLlib implements A2C and A3C using SyncSamplesOptimizer and AsyncGradientsOptimizer respectively for policy optimization. These algorithms scale to up to 16-32 worker processes depending on the environment. Both a TensorFlow (LSTM), and PyTorch version are available.
A2C also supports microbatching (i.e., gradient accumulation), which can be enabled by setting the ``microbatch_size`` config. Microbatching allows for training with a ``train_batch_size`` much larger than GPU memory. See also the `microbatch optimizer implementation <https://github.com/ray-project/ray/blob/master/rllib/optimizers/microbatch_optimizer.py>`__.
DDPG is implemented similarly to DQN (below). The algorithm can be scaled by increasing the number of workers, switching to AsyncGradientsOptimizer, or using Ape-X. The improvements from `TD3 <https://spinningup.openai.com/en/latest/algorithms/td3.html>`__ are available as ``TD3``.
RLlib DQN is implemented using the SyncReplayOptimizer. The algorithm can be scaled by increasing the number of workers, using the AsyncGradientsOptimizer for async DQN, or using Ape-X. Memory usage is reduced by compressing samples in the replay buffer with LZ4. All of the DQN improvements evaluated in `Rainbow <https://arxiv.org/abs/1710.02298>`__ are available, though not all are enabled by default. See also how to use `parametric-actions in DQN <rllib-models.html#variable-length-parametric-action-spaces>`__.
`[paper] <https://papers.nips.cc/paper/1713-policy-gradient-methods-for-reinforcement-learning-with-function-approximation.pdf>`__`[implementation] <https://github.com/ray-project/ray/blob/master/rllib/agents/pg/pg.py>`__ We include a vanilla policy gradients implementation as an example algorithm in both TensorFlow and PyTorch. This is usually outperformed by PPO.
PPO's clipped objective supports multiple SGD passes over the same batch of experiences. RLlib's multi-GPU optimizer pins that data in GPU memory to avoid unnecessary transfers from host memory, substantially improving performance over a naive implementation. RLlib's PPO scales out using multiple workers for experience collection, and also with multiple GPUs for SGD.
RLlib's multi-GPU PPO scales to multiple GPUs and hundreds of CPUs on solving the Humanoid-v1 task. Here we compare against a reference MPI-based implementation.
RLlib's soft-actor critic implementation is ported from the `official SAC repo <https://github.com/rail-berkeley/softlearning>`__ to better integrate with RLlib APIs. Note that SAC has two fields to configure for custom models: ``policy_model`` and ``Q_model``, and currently has no support for non-continuous action distributions.
ARS is a random search method for training linear policies for continuous control problems. Code here is adapted from https://github.com/modestyachts/ARS to integrate with RLlib APIs.
`[paper] <https://arxiv.org/abs/1803.11485>`__`[implementation] <https://github.com/ray-project/ray/blob/master/rllib/agents/qmix/qmix.py>`__ Q-Mix is a specialized multi-agent algorithm. Code here is adapted from https://github.com/oxwhirl/pymarl_alpha to integrate with RLlib multi-agent APIs. To use Q-Mix, you must specify an agent `grouping <rllib-env.html#grouping-agents>`__ in the environment (see the `two-step game example <https://github.com/ray-project/ray/blob/master/rllib/examples/twostep_game.py>`__). Currently, all agents in the group must be homogeneous. The algorithm can be scaled by increasing the number of workers or using Ape-X.
`[paper] <https://arxiv.org/abs/1706.02275>`__`[implementation] <https://github.com/ray-project/ray/blob/master/rllib/contrib/maddpg/maddpg.py>`__ MADDPG is a specialized multi-agent algorithm. Code here is adapted from https://github.com/openai/maddpg to integrate with RLlib multi-agent APIs. Please check `wsjeon/maddpg-rllib <https://github.com/wsjeon/maddpg-rllib>`__ for examples and more information.
**MADDPG-specific configs** (see also `common configs <rllib-training.html#common-parameters>`__):
Tuned examples: `Multi-Agent Particle Environment <https://github.com/wsjeon/maddpg-rllib/tree/master/plots>`__, `Two-step game <https://github.com/ray-project/ray/blob/master/rllib/examples/twostep_game.py>`__
`[paper] <http://papers.nips.cc/paper/7866-exponentially-weighted-imitation-learning-for-batched-historical-data>`__`[implementation] <https://github.com/ray-project/ray/blob/master/rllib/agents/marwil/marwil.py>`__ MARWIL is a hybrid imitation learning and policy gradient algorithm suitable for training on batched historical data. When the ``beta`` hyperparameter is set to zero, the MARWIL objective reduces to vanilla imitation learning. MARWIL requires the `offline datasets API <rllib-offline.html>`__ to be used.
`[paper] <https://arxiv.org/abs/1712.01815>`__`[implementation] <https://github.com/ray-project/ray/blob/master/rllib/contrib/alpha_zero>`__ AlphaZero is an RL agent originally designed for two-player games. This version adapts it to handle single player games. The code can be used with the SyncSamplesOptimizer as well as with a modified version of the SyncReplayOptimizer, and it scales to any number of workers. It also implements the ranked rewards `(R2) <https://arxiv.org/abs/1807.01672>`__ strategy to enable self-play even in the one-player setting. The code is mainly purposed to be used for combinatorial optimization.
