https://github.com/tgangwani/selfimitationdiverse
Tensorflow code for "Learning Self-Imitating Diverse Policies" (ICLR 2019)
https://github.com/tgangwani/selfimitationdiverse
deep-learning imitation-learning reinforcement-learning tensorflow
Last synced: 9 months ago
JSON representation
Tensorflow code for "Learning Self-Imitating Diverse Policies" (ICLR 2019)
- Host: GitHub
- URL: https://github.com/tgangwani/selfimitationdiverse
- Owner: tgangwani
- Created: 2019-05-03T22:49:57.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2020-11-26T20:50:07.000Z (over 5 years ago)
- Last Synced: 2025-04-13T16:48:02.359Z (about 1 year ago)
- Topics: deep-learning, imitation-learning, reinforcement-learning, tensorflow
- Language: Python
- Homepage:
- Size: 27.3 KB
- Stars: 19
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
This repo contains code for our paper [Learning Self-Imitating Diverse Policies](https://arxiv.org/abs/1805.10309) published at ICLR 2019.
The code was tested with the following packages:
* python 3.5.2
* tensorflow 1.4.0
* gym 0.9.2
## Running command
To train a self-imitation agent in an episodic reward environment, use:
```
python main.py --env_id HalfCheetah-v1 --seed=$(echo $RANDOM) --mu=0.8 --episodic
```
The parameter 'mu' is as defined in the paper (Equation 5.)
## SVPG for diverse multi-agent training
This functionality is provided as part of a separate codebase. Please use the code [here](https://github.com/tgangwani/QDAgents) with the following configuration in the file __default_config.yaml__: divergence: js, dre_type: nce
## Credits
The code is built on, and uses many utils from [OpenAI baselines](https://github.com/openai/baselines)