Skip to main content

Disributed RL implementations with ray and pytorch.

Project description

PyTorchRL: A PyTorch library for reinforcement learning

Deep Reinforcement learning (DRL) has been very successful in recent years but current methods still require vast amounts of data to solve non-trivial environments. Scaling to solve more complex tasks requires frameworks that are flexible enough to allow prototyping and testing of new ideas, yet avoiding the impractically slow experimental turnaround times associated to single-threaded implementations. PyTorchRL is a pytorch-based library for DRL that allows to easily assemble RL agents using a set of core reusable and easily extendable sub-modules as building blocks. To reduce training times, PyTorchRL allows scaling agents with a parameterizable component called Scheme, that permits to define distributed architectures with great flexibility by specifying which operations should be decoupled, which should be parallelized, and how parallel tasks should be synchronized.

Installation

    conda create -y -n pytorchrl
    conda activate pytorchrl

    conda install pytorch torchvision cudatoolkit -c pytorch

    pip install pytorchrl gym[atari,accept-rom-license]==0.22.0 wandb opencv-python hydra-core

Documentation

PyTorchRL documentation can be found here.

Citing PyTorchRL

Here is the paper

@misc{bou2021pytorchrl,
      title={PyTorchRL: Modular and Distributed Reinforcement Learning in PyTorch},
      author={Albert Bou and Gianni De Fabritiis},
      year={2021},
      eprint={2007.02622},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Project details


Release history Release notifications | RSS feed

Download files

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

Source Distribution

pytorchrl-3.2.15.tar.gz (144.5 kB view details)

Uploaded Source

Built Distribution

pytorchrl-3.2.15-py3-none-any.whl (246.4 kB view details)

Uploaded Python 3

File details

Details for the file pytorchrl-3.2.15.tar.gz.

File metadata

  • Download URL: pytorchrl-3.2.15.tar.gz
  • Upload date:
  • Size: 144.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.3

File hashes

Hashes for pytorchrl-3.2.15.tar.gz
Algorithm Hash digest
SHA256 0239717b5f944390c2727457521b952de5e75a6702f83601e8407bdc14e2e7ac
MD5 b47e8028ee578cacd04f299634419e6c
BLAKE2b-256 efe6a2266d57ff99e473a21ae2fb6427411d584deb846b541c1ddd610d33b00f

See more details on using hashes here.

File details

Details for the file pytorchrl-3.2.15-py3-none-any.whl.

File metadata

  • Download URL: pytorchrl-3.2.15-py3-none-any.whl
  • Upload date:
  • Size: 246.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.11.3

File hashes

Hashes for pytorchrl-3.2.15-py3-none-any.whl
Algorithm Hash digest
SHA256 a1a9a8fe7924b9ba0e47682cde418a42daa491f3fbeae5bb2b75cb59a1ac2818
MD5 5d16060d15dce12acdde01000e744c20
BLAKE2b-256 55b9e70b4edb8141003842648825f2e420c46d140b7147273f63035a720aeb30

See more details on using hashes here.

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