Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).
| Date | Stars |
|---|---|
| 2026-07-24 | 3903 |
| 2026-07-25 | 3903 |
| 2026-07-28 | 3903 |
| 2026-07-30 | 3903 |
| 2026-07-31 | 3905 |
| 2026-08-01 | 3904 |
| 2026-08-02 | 3905 |
| 2026-08-03 | 3904 |
| 2026-08-06 | 3904 |
Today
— stars today
This week
+1 stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.03%/day
# pytorch-a2c-ppo-acktr
## Update (April 12th, 2021)
PPO is great, but [Soft Actor Critic](https://arxiv.org/abs/1812.05905) can be better for many continuous control tasks. Please check out [my new RL](http://github.com/ikostrikov/jax-rl) repository in jax.
## Please use hyper parameters from this readme. With other hyper parameters things might not work (it's RL after all)!
This is a PyTorch implementation of
* Advantage Actor Critic (A2C), a synchronous deterministic version of [A3C](https://arxiv.org/pdf/1602.01783v1.pdf)
* Proximal Policy Optimization [PPO](https://arxiv.org/pdf/1707.06347.pdf)
* Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation [ACKTR](https://arxiv.org/abs/1708.05144)
* Generative Adversarial Imitation Learning [GAIL](https://arxiv.org/abs/1606.03476)
Also see the OpenAI posts: [A2C/ACKTR](https://blog.openai.com/baselines-acktr-a2c/) and [PPO](https://blog.openai.com/openai-baselines-ppo/) for more information.
This implementation is inspired by the OpenAI baselines for [A2C](https://github.com/openai/baselines/tree/master/baselines/a2c), [ACKTR](https://github.com/openai/baselines/tree/master/baselines/acktr) and [PPO](https://github.com/openai/baselines/tree/master/baselines/ppo1). It uses the same hyper parameters and the model since they were well tuned for Atari games.
Please use this bibtex if you want to cite this repository in your publications:
@misc{pytorchrl,
author = {Kostrikov, Ilya},
title = {PyTorch Implementations of Reinforcement Learning Algorithms},
year = {2018},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/ikostrikov/pytorch-a2c-ppo-acktr-gail}},
}
## Supported (and tested) environments (via [OpenAI Gym](https://gym.openai.com))
* [Atari Learning Environment](https://github.com/mgbellemare/Arcade-Learning-Environment)
* [MuJoCo](http://mujoco.org)
* [PyBullet](http://pybullet.org) (including Racecar, Minitaur and Kuka)
* [DeepMind Control Suite](https://github.com/deepmind/dm_control) (via [dm_control2gym](https://github.com/martinseilair/dm_control2gym))
I highly recommend PyBullet as a free open source alternative to MuJoCo for continuous control tasks.
All environments are operated using exactly the same Gym interface. See their documentations for a comprehensive list.
To use the DeepMind Control Suite environments, set the flag `--env-name dm.<domain_name>.<task_name>`, where `domain_name` and `task_name` are the name of a domain (e.g. `hopper`) and a task within that domain (e.g. `stand`) from the DeepMind Control Suite. Refer to their repo and their [tech report](https://arxiv.org/abs/1801.00690) for a full list of available domains and tasks. Other than setting the task, the API for interacting with the environment is exactly the same as for all the Gym environments thanks to [dm_control2gym](https://github.com/martinseilair/dm_control2gym).
## Requirements
* Python 3 (it might work with Python 2, but I didn't test it)
* [PyTorch](http://pytorch.org/)
* [Stable baselines3](https://github.com/DLR-RM/stable-baselines3)
In order to install requirements, follow:
```bash
# PyTorch
conda install pytorch torchvision -c soumith
# Other requirements
pip install -r requirements.txt
# Gym Atari
conda install -c conda-forge gym-atari
```
## Contributions
Contributions are very welcome. If you know how to make this code better, please open an issue. If you want to submit a pull request, please open an issue first. Also see a todo list below.
Also I'm searching for volunteers to run all experiments on Atari and MuJoCo (with multiple random seeds).
## Disclaimer
It's extremely difficult to reproduce results for Reinforcement Learning methods. See ["Deep Reinforcement Learning that Matters"](https://arxiv.org/abs/1709.06560) for more information. I tried to reproduce OpenAI results as closely as possible. However, majors differeExcerpt of 6,460 characters
Read on GitHub
201
15
Erik Wijmans
6
Maxime Chevalier-Boisvert · Canada
5
5
4
Ilya Kostrikov
3
Costa Huang · @periodic · United States
3
2
Ross Wightman · Canada
2
Rana Hanocka · Assistant Professor
2
Wei Yang · NVIDIA · United States
2
Haresh Karnan · The University of Texas at Austin · United States
2
2
2
2
Mihir Rana · Meta · United States
2
Alok Singh · United States
1
1
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:5867ddba36cb91b3, topic:reinforcement-learning, topic:deep-reinforcement-learning, desc:reinforcement learning
matched fp:5867ddba36cb91b3, topic:deep-learning, topic:pytorch