This repository contains the official implementation for the paper Augmented Bayesien Policy Search accepted at ICLR 2024.

• Installation
• Usage
• License

Create new virtual environment :

python -m venv /path/to/new/virtual/environment

Install requirements file :

python -m venv /path/to/new/virtual/environment

Install jax separately depending on your hardware (a GPU is definetely recommended):

CPU: pip install -U "jax[cpu]

GPU:

pip install -U "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

You need to specify a wandb api key for the environment variables at each script.

To run the ARS baseline just run :

python run_ars.py

To run MPD :

python run_mpd.py --algo_name mpd

You can an algorithm with parameters different from the default configs e.g :

python run_mpd.py --algo_name abs --reset_critic False --n_critics 10

This project is licensed under the MIT License.