https://github.com/lebellig/dueling-network-architectures
https://github.com/lebellig/dueling-network-architectures
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/lebellig/dueling-network-architectures
- Owner: lebellig
- License: mit
- Created: 2021-12-29T16:36:30.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2022-01-28T15:26:19.000Z (over 4 years ago)
- Last Synced: 2025-09-08T12:47:50.246Z (11 months ago)
- Language: Jupyter Notebook
- Size: 397 KB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Dueling Network Architectures for Reinforcement Learning
## Project structure
The two main experiments made are located in the `notebooks/` folder (for dueling network architectures and single stream model for Q-learning) and you can watch the videos of the spaceship landing on the moon in the `videos/`.
All models and their blocks are located in the `src/dueling_archs/models` folder and the corresponding losses in the `src/dueling_archs/losses`.
https://user-images.githubusercontent.com/47056792/151573866-5c0ea2e1-12bc-40e5-b675-65967e84a150.mp4
Run tests with *pytest* from `test/`.
## Installation
1. Clone this repository:
```bash
git clone https://github.com/gle-bellier/dueling-network-architectures.git
```
2. Install requirements:
```bash
cd expressive-perf
pip install -r requirements.txt
```
3. Install package locally:
```bash
pip install -e .
```