A-LIX on DMC

PyTorch implementation of Adaptive Local Signal Mixing from Stabilizing Off-Policy Deep Reinforcement Learning from Pixels. This repository can be used to reproduce DeepMind Control experiments.

For further details see our ICML 2022 paper:

Instructions

Install MuJoCo

Install dependencies with conda:

conda env create -f conda_env.yml
conda activate drqv2

Train an agent/collect results

Use Hydra configuration files (provided in the cfgs folder), specifying algo and env, representing the algorithm and environment configurations.

E.g.:

python train.py algo=ALIX task=quadruped_walk

You can monitor with tensorboard by running:

tensorboard --logdir exp_local

Extend/contact

The main classes/functions used for A-LIX are located in the analysis_* files.

For any queries/questions, feel free to raise an issue and/or get in contact with Edoardo Cetin or Philip J. Ball.

To cite our work, use:

@inproceedings{cetin2022stabilizing,
  title={Stabilizing Off-Policy Deep Reinforcement Learning from Pixels},
  author={Cetin, Edoardo and Ball, Philip J and Roberts, Stephen and Celiktutan, Oya},
  booktitle={International Conference on Machine Learning},
  pages={2784--2810},
  year={2022},
  organization={PMLR}
}

Acknowledgements

We would like to thank Denis Yarats for open-sourcing the DrQv2 codebase. Our implementation builds on top of their repository.