https://github.com/ikostrikov/pytorch-trpo

PyTorch implementation of Trust Region Policy Optimization
https://github.com/ikostrikov/pytorch-trpo

continuous-control deep-learning deep-reinforcement-learning mujoco pytorch reinforcement-learning trpo trust-region-policy-optimization

Last synced: about 23 hours ago
JSON representation

PyTorch implementation of Trust Region Policy Optimization

• Host: GitHub
• URL: https://github.com/ikostrikov/pytorch-trpo
• Owner: ikostrikov
• License: mit
• Created: 2017-05-26T12:33:18.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2018-09-13T13:59:57.000Z (over 6 years ago)
• Last Synced: 2025-01-28T21:09:25.300Z (9 days ago)
• Topics: continuous-control, deep-learning, deep-reinforcement-learning, mujoco, pytorch, reinforcement-learning, trpo, trust-region-policy-optimization
• Language: Python
• Homepage:
• Size: 9.77 KB
• Stars: 435
• Watchers: 12
• Forks: 91
• Open Issues: 7
• Metadata Files:
  • Readme: README.md
  • License: LICENSE.md

Awesome Lists containing this project

README

        
# PyTorch implementation of TRPO

Try my implementation of [PPO](github.com/ikostrikov/pytorch-a2c-ppo-acktr/) (aka newer better variant of TRPO), unless you need to you TRPO for some specific reasons.

##

This is a PyTorch implementation of ["Trust Region Policy Optimization (TRPO)"](https://arxiv.org/abs/1502.05477).

This is code mostly ported from [original implementation by John Schulman](https://github.com/joschu/modular_rl). In contrast to [another implementation of TRPO in PyTorch](https://github.com/mjacar/pytorch-trpo), this implementation uses exact Hessian-vector product instead of finite differences approximation.

## Contributions

Contributions are very welcome. If you know how to make this code better, don't hesitate to send a pull request.

## Usage

```

python main.py --env-name "Reacher-v1"

```

## Recommended hyper parameters

InvertedPendulum-v1: 5000

Reacher-v1, InvertedDoublePendulum-v1: 15000

HalfCheetah-v1, Hopper-v1, Swimmer-v1, Walker2d-v1: 25000

Ant-v1, Humanoid-v1: 50000

## Results

More or less similar to the original code. Coming soon.

## Todo

- [ ] Plots.

- [ ] Collect data in multiple threads.