DQN implementation for checking baselines.
- Python3
- tensorflow
- gym[atari]
- opencv-python
- git+https://github.com/imai-laboratory/rlsaber
$ python train.py [--render]
$ python train.py [--render] [--load {path of models}] --demo
- Use
***Deterministic-v4
instead of***-v0
for environemts because in default, environments sample {2, 3, 4} frames uniformaly. This affects negatively performance because of non-deterministic frame skipping.
This baseline is inspired by following projects.