https://github.com/junming-yang/moan
Model-based Offline Policy Optimization with Adversarial Network
https://github.com/junming-yang/moan
Last synced: 12 months ago
JSON representation
Model-based Offline Policy Optimization with Adversarial Network
- Host: GitHub
- URL: https://github.com/junming-yang/moan
- Owner: junming-yang
- License: mit
- Created: 2023-07-30T07:35:22.000Z (almost 3 years ago)
- Default Branch: v1
- Last Pushed: 2023-11-09T13:06:06.000Z (over 2 years ago)
- Last Synced: 2025-01-16T14:29:35.546Z (over 1 year ago)
- Size: 6.69 MB
- Stars: 5
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Overview
This is an official implementation of the offline model-based RL algorithm [MOAN](https://arxiv.org/abs/2309.02157) all by pytorch.
# Dependencies
- MuJoCo 2.0
- Gym 0.22.0
- D4RL
- PyTorch 1.8+
# Usage
## Train
```
# for hopper-medium-replay-v2 task
python3 train.py --task "hopper-medium-replay-v2" --rollout-length 5 --seed 50 --d-coeff 0.1
# for halfcheetah-medium-replay-v2 task
python3 train.py --task "hopper-medium-replay-v2" --rollout-length 5 --seed 50 --d-coeff 0.1
# for walker2d-medium-replay-v2 task
python3 train.py --task "hopper-medium-replay-v2" --rollout-length 5 --seed 50 --d-coeff 0.1
```
# Reference
If you find this code useful, please reference in our paper:
```
@article{yang2023model,
title={Model-based offline policy optimization with adversarial network},
author={Yang, Junming and Chen, Xingguo and Wang, Shengyuan and Zhang, Bolei},
journal={arXiv preprint arXiv:2309.02157},
year={2023}
}
```