Pytorch Implementation of Rainbow
This repo is a partial implementation of the Rainbow agent published by researchers from DeepMind. The implementation is efficient and of high quality. It trains at a speed of 350 frames/s on a PC with a 3.5GHz CPU and GTX1080 GPU.
Rainbow is a deep Q learning based agent that combines a bunch of existing techiques such as dueling dqn, distributional dqn, etc. This repo currenly implemented the following dqn variants:
and it will need the following extensions to become a full "Rainbow":
- Multi-step learning
- Priority Replay
The hyperparameters in this repo follows the ones described in Rainbow paper as close as possible. However, there may still be some differences due to misunderstanding.
DQN agent often takes days to train. For sanity check, we can train a agent to play a simple game "boxing". Follwing is the learning curve of a dueling double dqn trained on boxing.
The agent almost solves boxing after around 12M frames, which is a good sign that the implementation is working.
To test the distributional DQN and Noisy Net, the agent is trained on "breakout" since distributional DQN performs significantly better than others on this game, reaching >400 scores rapidly while other DQN methods struggle to do so.
From the figure we see that the agent can reach >400 scores very rapidly and steadily. Note that the publicly reported numbers on papers are produced by training the agent for 200M frames while here it trains only for 50M frames due to computation cost.
Figures here are smoothed.
We plan to implement multi-step learing and priority replay. Also, the current implementation uses a simple wrapper on the Arcade Learning Enviroment. We may want to shift to OpenAI gym for better visualization and video recording. On top of Rainbow, it will also be interesting to include other new techniques, such as Distributional RL with Quantile Regression.
Contributions and bug-catchings are welcome!