TorchRL

Documentation | TensorDict |
Features | Examples, tutorials and demos | Citation | Installation |
Asking a question | Contributing

TorchRL is an open-source Reinforcement Learning (RL) library for PyTorch.

Key features

• 🐍 Python-first: Designed with Python as the primary language for ease of use and flexibility
• ⏱️ Efficient: Optimized for performance to support demanding RL research applications
• 🧮 Modular, customizable, extensible: Highly modular architecture allows for easy swapping, transformation, or creation of new components
• 📚 Documented: Thorough documentation ensures that users can quickly understand and utilize the library
• ✅ Tested: Rigorously tested to ensure reliability and stability
• ⚙️ Reusable functionals: Provides a set of highly reusable functions for cost functions, returns, and data processing

Design Principles

• 🔥 Aligns with PyTorch ecosystem: Follows the structure and conventions of popular PyTorch libraries
  (e.g., dataset pillar, transforms, models, data utilities)
• ➖ Minimal dependencies: Only requires Python standard library, NumPy, and PyTorch; optional dependencies for
  common environment libraries (e.g., OpenAI Gym) and datasets (D4RL, OpenX…)

Read the full paper for a more curated description of the library.

Getting started

Check our Getting Started tutorials for quickly ramp up with the basic
features of the library!

Documentation and knowledge base

The TorchRL documentation can be found here.
It contains tutorials and the API reference.

TorchRL also provides a RL knowledge base to help you debug your code, or simply
learn the basics of RL. Check it out here.

We have some introductory videos for you to get to know the library better, check them out:

• TalkRL podcast
• TorchRL intro at PyTorch day 2022
• PyTorch 2.0 Q&A: TorchRL

Spotlight publications

TorchRL being domain-agnostic, you can use it across many different fields. Here are a few examples:

• ACEGEN: Reinforcement Learning of Generative Chemical Agents
  for Drug Discovery
• BenchMARL: Benchmarking Multi-Agent Reinforcement Learning
• BricksRL: A Platform for Democratizing Robotics and Reinforcement Learning
  Research and Education with LEGO
• OmniDrones: An Efficient and Flexible Platform for Reinforcement Learning in Drone Control
• RL4CO: an Extensive Reinforcement Learning for Combinatorial Optimization Benchmark
• Robohive: A unified framework for robot learning

Writing simplified and portable RL codebase with TensorDict

RL algorithms are very heterogeneous, and it can be hard to recycle a codebase
across settings (e.g. from online to offline, from state-based to pixel-based
learning).
TorchRL solves this problem through TensorDict,
a convenient data structure⁽¹⁾ that can be used to streamline one’s
RL codebase.
With this tool, one can write a complete PPO training script in less than 100
lines of code!

import torch
from tensordict.nn import TensorDictModule
from tensordict.nn.distributions import NormalParamExtractor
from torch import nn

from torchrl.collectors import SyncDataCollector
from torchrl.data.replay_buffers import TensorDictReplayBuffer, \
  LazyTensorStorage, SamplerWithoutReplacement
from torchrl.envs.libs.gym import GymEnv
from torchrl.modules import ProbabilisticActor, ValueOperator, TanhNormal
from torchrl.objectives import ClipPPOLoss
from torchrl.objectives.value import GAE

env = GymEnv("Pendulum-v1") 
model = TensorDictModule(
  nn.Sequential(
      nn.Linear(3, 128), nn.Tanh(),
      nn.Linear(128, 128), nn.Tanh(),
      nn.Linear(128, 128), nn.Tanh(),
      nn.Linear(128, 2),
      NormalParamExtractor()
  ),
  in_keys=["observation"],
  out_keys=["loc", "scale"]
)
critic = ValueOperator(
  nn.Sequential(
      nn.Linear(3, 128), nn.Tanh(),
      nn.Linear(128, 128), nn.Tanh(),
      nn.Linear(128, 128), nn.Tanh(),
      nn.Linear(128, 1),
  ),
  in_keys=["observation"],
)
actor = ProbabilisticActor(
  model,
  in_keys=["loc", "scale"],
  distribution_class=TanhNormal,
  distribution_kwargs={"low": -1.0, "high": 1.0},
  return_log_prob=True
  )
buffer = TensorDictReplayBuffer(
  storage=LazyTensorStorage(1000),
  sampler=SamplerWithoutReplacement(),
  batch_size=50,
  )
collector = SyncDataCollector(
  env,
  actor,
  frames_per_batch=1000,
  total_frames=1_000_000,
)
loss_fn = ClipPPOLoss(actor, critic)
adv_fn = GAE(value_network=critic, average_gae=True, gamma=0.99, lmbda=0.95)
optim = torch.optim.Adam(loss_fn.parameters(), lr=2e-4)

for data in collector:  # collect data
  for epoch in range(10):
      adv_fn(data)  # compute advantage
      buffer.extend(data)
      for sample in buffer:  # consume data
          loss_vals = loss_fn(sample)
          loss_val = sum(
              value for key, value in loss_vals.items() if
              key.startswith("loss")
              )
          loss_val.backward()
          optim.step()
          optim.zero_grad()
  print(f"avg reward: {data['next', 'reward'].mean().item(): 4.4f}")

Here is an example of how the environment API
relies on tensordict to carry data from one function to another during a rollout
execution:

TensorDict makes it easy to re-use pieces of code across environments, models and
algorithms.

- obs, done = env.reset()
+ tensordict = env.reset()
policy = SafeModule(
    model,
    in_keys=["observation_pixels", "observation_vector"],
    out_keys=["action"],
)
out = []
for i in range(n_steps):
-     action, log_prob = policy(obs)
-     next_obs, reward, done, info = env.step(action)
-     out.append((obs, next_obs, action, log_prob, reward, done))
-     obs = next_obs
+     tensordict = policy(tensordict)
+     tensordict = env.step(tensordict)
+     out.append(tensordict)
+     tensordict = step_mdp(tensordict)  # renames next_observation_* keys to observation_*
- obs, next_obs, action, log_prob, reward, done = [torch.stack(vals, 0) for vals in zip(*out)]
+ out = torch.stack(out, 0)  # TensorDict supports multiple tensor operations

Using this, TorchRL abstracts away the input / output signatures of the modules, env,
collectors, replay buffers and losses of the library, allowing all primitives
to be easily recycled across settings.

- for i, (obs, next_obs, action, hidden_state, reward, done) in enumerate(collector):
+ for i, tensordict in enumerate(collector):
-     replay_buffer.add((obs, next_obs, action, log_prob, reward, done))
+     replay_buffer.add(tensordict)
    for j in range(num_optim_steps):
-         obs, next_obs, action, hidden_state, reward, done = replay_buffer.sample(batch_size)
-         loss = loss_fn(obs, next_obs, action, hidden_state, reward, done)
+         tensordict = replay_buffer.sample(batch_size)
+         loss = loss_fn(tensordict)
        loss.backward()
        optim.step()
        optim.zero_grad()

TensorDict supports multiple tensor operations on its device and shape
(the shape of TensorDict, or its batch size, is the common arbitrary N first dimensions of all its contained tensors):

# stack and cat
tensordict = torch.stack(list_of_tensordicts, 0)
tensordict = torch.cat(list_of_tensordicts, 0)
# reshape
tensordict = tensordict.view(-1)
tensordict = tensordict.permute(0, 2, 1)
tensordict = tensordict.unsqueeze(-1)
tensordict = tensordict.squeeze(-1)
# indexing
tensordict = tensordict[:2]
tensordict[:, 2] = sub_tensordict
# device and memory location
tensordict.cuda()
tensordict.to("cuda:1")
tensordict.share_memory_()

TensorDict comes with a dedicated tensordict.nn
module that contains everything you might need to write your model with it.
And it is functorch and torch.compile compatible!

transformer_model = nn.Transformer(nhead=16, num_encoder_layers=12)
+ td_module = SafeModule(transformer_model, in_keys=["src", "tgt"], out_keys=["out"])
src = torch.rand((10, 32, 512))
tgt = torch.rand((20, 32, 512))
+ tensordict = TensorDict({"src": src, "tgt": tgt}, batch_size=[20, 32])
- out = transformer_model(src, tgt)
+ td_module(tensordict)
+ out = tensordict["out"]

encoder_module = TransformerEncoder(...)
encoder = TensorDictSequential(encoder_module, in_keys=["src", "src_mask"], out_keys=["memory"])
decoder_module = TransformerDecoder(...)
decoder = TensorDictModule(decoder_module, in_keys=["tgt", "memory"], out_keys=["output"])
transformer = TensorDictSequential(encoder, decoder)
assert transformer.in_keys == ["src", "src_mask", "tgt"]
assert transformer.out_keys == ["memory", "output"]

transformer.select_subsequence(out_keys=["memory"])  # returns the encoder
transformer.select_subsequence(in_keys=["tgt", "memory"])  # returns the decoder

Check TensorDict tutorials to
learn more!

Features

• A common interface for environments
  which supports common libraries (OpenAI gym, deepmind control lab, etc.)⁽¹⁾ and state-less execution
  (e.g. Model-based environments).
  The batched environments containers allow parallel execution⁽²⁾.
  A common PyTorch-first class of tensor-specification class is also provided.
  TorchRL’s environments API is simple but stringent and specific. Check the
  documentation
  and tutorial to learn more!

  Code

  env_make = lambda: GymEnv("Pendulum-v1", from_pixels=True)
  env_parallel = ParallelEnv(4, env_make)  # creates 4 envs in parallel
  tensordict = env_parallel.rollout(max_steps=20, policy=None)  # random rollout (no policy given)
  assert tensordict.shape == [4, 20]  # 4 envs, 20 steps rollout
  env_parallel.action_spec.is_in(tensordict["action"])  # spec check returns True
  

• multiprocess and distributed data collectors⁽²⁾
  that work synchronously or asynchronously.
  Through the use of TensorDict, TorchRL’s training loops are made very similar
  to regular training loops in supervised
  learning (although the “dataloader” – read data collector – is modified on-the-fly):

  Code

  env_make = lambda: GymEnv("Pendulum-v1", from_pixels=True)
  collector = MultiaSyncDataCollector(
      [env_make, env_make],
      policy=policy,
      devices=["cuda:0", "cuda:0"],
      total_frames=10000,
      frames_per_batch=50,
      ...
  )
  for i, tensordict_data in enumerate(collector):
      loss = loss_module(tensordict_data)
      loss.backward()
      optim.step()
      optim.zero_grad()
      collector.update_policy_weights_()
  

  Check our distributed collector examples to
  learn more about ultra-fast data collection with TorchRL.

• efficient⁽²⁾ and generic⁽¹⁾ replay buffers with modularized storage:

  Code

  storage = LazyMemmapStorage(  # memory-mapped (physical) storage
      cfg.buffer_size,
      scratch_dir="/tmp/"
  )
  buffer = TensorDictPrioritizedReplayBuffer(
      alpha=0.7,
      beta=0.5,
      collate_fn=lambda x: x,
      pin_memory=device != torch.device("cpu"),
      prefetch=10,  # multi-threaded sampling
      storage=storage
  )
  

  Replay buffers are also offered as wrappers around common datasets for offline RL:

  Code

  from torchrl.data.replay_buffers import SamplerWithoutReplacement
  from torchrl.data.datasets.d4rl import D4RLExperienceReplay
  data = D4RLExperienceReplay(
      "maze2d-open-v0",
      split_trajs=True,
      batch_size=128,
      sampler=SamplerWithoutReplacement(drop_last=True),
  )
  for sample in data:  # or alternatively sample = data.sample()
      fun(sample)
  

• cross-library environment transforms⁽¹⁾,
  executed on device and in a vectorized fashion⁽²⁾,
  which process and prepare the data coming out of the environments to be used by the agent:

  Code

  env_make = lambda: GymEnv("Pendulum-v1", from_pixels=True)
  env_base = ParallelEnv(4, env_make, device="cuda:0")  # creates 4 envs in parallel
  env = TransformedEnv(
      env_base,
      Compose(
          ToTensorImage(),
          ObservationNorm(loc=0.5, scale=1.0)),  # executes the transforms once and on device
  )
  tensordict = env.reset()
  assert tensordict.device == torch.device("cuda:0")
  

  Other transforms include: reward scaling (RewardScaling), shape operations (concatenation of tensors, unsqueezing etc.), concatenation of
  successive operations (CatFrames), resizing (Resize) and many more.

  Unlike other libraries, the transforms are stacked as a list (and not wrapped in each other), which makes it
  easy to add and remove them at will:

  env.insert_transform(0, NoopResetEnv())  # inserts the NoopResetEnv transform at the index 0
  

  Nevertheless, transforms can access and execute operations on the parent environment:

  transform = env.transform[1]  # gathers the second transform of the list
  parent_env = transform.parent  # returns the base environment of the second transform, i.e. the base env + the first transform
  

• various tools for distributed learning (e.g. memory mapped tensors)⁽²⁾;

• various architectures and models (e.g. actor-critic)⁽¹⁾:

  Code

  # create an nn.Module
  common_module = ConvNet(
      bias_last_layer=True,
      depth=None,
      num_cells=[32, 64, 64],
      kernel_sizes=[8, 4, 3],
      strides=[4, 2, 1],
  )
  # Wrap it in a SafeModule, indicating what key to read in and where to
  # write out the output
  common_module = SafeModule(
      common_module,
      in_keys=["pixels"],
      out_keys=["hidden"],
  )
  # Wrap the policy module in NormalParamsWrapper, such that the output
  # tensor is split in loc and scale, and scale is mapped onto a positive space
  policy_module = SafeModule(
      NormalParamsWrapper(
          MLP(num_cells=[64, 64], out_features=32, activation=nn.ELU)
      ),
      in_keys=["hidden"],
      out_keys=["loc", "scale"],
  )
  # Use a SafeProbabilisticTensorDictSequential to combine the SafeModule with a
  # SafeProbabilisticModule, indicating how to build the
  # torch.distribution.Distribution object and what to do with it
  policy_module = SafeProbabilisticTensorDictSequential(  # stochastic policy
      policy_module,
      SafeProbabilisticModule(
          in_keys=["loc", "scale"],
          out_keys="action",
          distribution_class=TanhNormal,
      ),
  )
  value_module = MLP(
      num_cells=[64, 64],
      out_features=1,
      activation=nn.ELU,
  )
  # Wrap the policy and value funciton in a common module
  actor_value = ActorValueOperator(common_module, policy_module, value_module)
  # standalone policy from this
  standalone_policy = actor_value.get_policy_operator()
  

• exploration wrappers and
  modules to easily swap between exploration and exploitation⁽¹⁾:

  Code

  policy_explore = EGreedyWrapper(policy)
  with set_exploration_type(ExplorationType.RANDOM):
      tensordict = policy_explore(tensordict)  # will use eps-greedy
  with set_exploration_type(ExplorationType.DETERMINISTIC):
      tensordict = policy_explore(tensordict)  # will not use eps-greedy
  

• A series of efficient loss modules
  and highly vectorized
  functional return and advantage
  computation.

  Code

  Loss modules

  from torchrl.objectives import DQNLoss
  loss_module = DQNLoss(value_network=value_network, gamma=0.99)
  tensordict = replay_buffer.sample(batch_size)
  loss = loss_module(tensordict)
  

  Advantage computation

  from torchrl.objectives.value.functional import vec_td_lambda_return_estimate
  advantage = vec_td_lambda_return_estimate(gamma, lmbda, next_state_value, reward, done, terminated)
  

• a generic trainer class⁽¹⁾ that
  executes the aforementioned training loop. Through a hooking mechanism,
  it also supports any logging or data transformation operation at any given
  time.

• various recipes to build models that
  correspond to the environment being deployed.

If you feel a feature is missing from the library, please submit an issue!
If you would like to contribute to new features, check our call for contributions and our contribution page.

Examples, tutorials and demos

A series of State-of-the-Art implementations are provided with an illustrative purpose:

** The number indicates expected speed-up compared to eager mode when executed on CPU. Numbers may vary depending on
architecture and device.

and many more to come!

Code examples displaying toy code snippets and training scripts are also available

• RLHF
• Memory-mapped replay buffers

Check the examples directory for more details
about handling the various configuration settings.

We also provide tutorials and demos that give a sense of
what the library can do.

Citation

If you’re using TorchRL, please refer to this BibTeX entry to cite this work:

@misc{bou2023torchrl,
      title={TorchRL: A data-driven decision-making library for PyTorch}, 
      author={Albert Bou and Matteo Bettini and Sebastian Dittert and Vikash Kumar and Shagun Sodhani and Xiaomeng Yang and Gianni De Fabritiis and Vincent Moens},
      year={2023},
      eprint={2306.00577},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

Installation

Create a conda environment where the packages will be installed.

conda create --name torch_rl python=3.9
conda activate torch_rl

PyTorch

Depending on the use of functorch that you want to make, you may want to
install the latest (nightly) PyTorch release or the latest stable version of PyTorch.
See here for a detailed list of commands,
including pip3 or other special installation instructions.

Torchrl

You can install the latest stable release by using

pip3 install torchrl

This should work on linux, Windows 10 and OsX (Intel or Silicon chips).
On certain Windows machines (Windows 11), one should install the library locally (see below).

The nightly build can be installed via

pip3 install torchrl-nightly

which we currently only ship for Linux and OsX (Intel) machines.
Importantly, the nightly builds require the nightly builds of PyTorch too.

To install extra dependencies, call

pip3 install "torchrl[atari,dm_control,gym_continuous,rendering,tests,utils,marl,open_spiel,checkpointing]"

or a subset of these.

One may also desire to install the library locally. Three main reasons can motivate this:

• the nightly/stable release isn’t available for one’s platform (eg, Windows 11, nightlies for Apple Silicon etc.);
• contributing to the code;
• install torchrl with a previous version of PyTorch (any version >= 2.0) (note that this should also be doable via a regular install followed
  by a downgrade to a previous pytorch version – but the C++ binaries will not be available so some feature will not work,
  such as prioritized replay buffers and the like.)

To install the library locally, start by cloning the repo:

git clone https://github.com/pytorch/rl

and don’t forget to check out the branch or tag you want to use for the build:

git checkout v0.4.0

Go to the directory where you have cloned the torchrl repo and install it (after
installing ninja)

cd /path/to/torchrl/
pip3 install ninja -U
python setup.py develop

One can also build the wheels to distribute to co-workers using

python setup.py bdist_wheel

Your wheels will be stored there ./dist/torchrl<name>.whl and installable via

pip install torchrl<name>.whl

Warning: Unfortunately, pip3 install -e . does not currently work. Contributions to help fix this are welcome!

On M1 machines, this should work out-of-the-box with the nightly build of PyTorch.
If the generation of this artifact in MacOs M1 doesn’t work correctly or in the execution the message
(mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64e')) appears, then try

ARCHFLAGS="-arch arm64" python setup.py develop

To run a quick sanity check, leave that directory (e.g. by executing cd ~/)
and try to import the library.

python -c "import torchrl"

This should not return any warning or error.

Optional dependencies

The following libraries can be installed depending on the usage one wants to
make of torchrl:

# diverse
pip3 install tqdm tensorboard "hydra-core>=1.1" hydra-submitit-launcher

# rendering
pip3 install "moviepy<2.0.0"

# deepmind control suite
pip3 install dm_control

# gym, atari games
pip3 install "gym[atari]" "gym[accept-rom-license]" pygame

# tests
pip3 install pytest pyyaml pytest-instafail

# tensorboard
pip3 install tensorboard

# wandb
pip3 install wandb

Troubleshooting

If a ModuleNotFoundError: No module named ‘torchrl._torchrl errors occurs (or
a warning indicating that the C++ binaries could not be loaded),
it means that the C++ extensions were not installed or not found.

• One common reason might be that you are trying to import torchrl from within the
  git repo location. The following code snippet should return an error if
  torchrl has not been installed in develop mode:

  cd ~/path/to/rl/repo
  python -c 'from torchrl.envs.libs.gym import GymEnv'
  

  If this is the case, consider executing torchrl from another location.
• If you’re not importing torchrl from within its repo location, it could be
  caused by a problem during the local installation. Check the log after the
  python setup.py develop. One common cause is a g++/C++ version discrepancy
  and/or a problem with the ninja library.
• If the problem persists, feel free to open an issue on the topic in the repo,
  we’ll make our best to help!
• On MacOs, we recommend installing XCode first.
  With Apple Silicon M1 chips, make sure you are using the arm64-built python
  (e.g. here).
  Running the following lines of code

  wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
  python collect_env.py
  

  should display

  OS: macOS *** (arm64)
  

  and not

  OS: macOS **** (x86_64)
  

Versioning issues can cause error message of the type undefined symbol
and such. For these, refer to the versioning issues document
for a complete explanation and proposed workarounds.

Asking a question

If you spot a bug in the library, please raise an issue in this repo.

If you have a more generic question regarding RL in PyTorch, post it on
the PyTorch forum.

Contributing

Internal collaborations to torchrl are welcome! Feel free to fork, submit issues and PRs.
You can checkout the detailed contribution guide here.
As mentioned above, a list of open contributions can be found in here.

Contributors are recommended to install pre-commit hooks (using pre-commit install). pre-commit will check for linting related issues when the code is committed locally. You can disable th check by appending -n to your commit command: git commit -m <commit message> -n

Disclaimer

This library is released as a PyTorch beta feature.
BC-breaking changes are likely to happen but they will be introduced with a deprecation
warranty after a few release cycles.

License

TorchRL is licensed under the MIT License. See LICENSE for details.