DQN implementation by Chainer.
Dependencies
- Chainer(>=1.15)
- gym(>=0.2.11)
python run.py
It iterators 5 episode. If you store the model on ./store directory, that is loaded.
You can use trained model that are located in trained_model directory (it is stored by Git LFS, storing latest 5 model).
Please copy it to /store directory then run script.
If you want to submit the result, set the option --submit <your api key>.
Below is the result by latest trained model.
OpenAI Gym/icoxfog417 algorithm
python run.py --train
If you use GPU, add --gpu option.
The trained model is stored in ./store directory. If you want to change the parameters, please modify run.py.
The training parameters is manged in DQNTrainer.