MBPO PyTorch

A PyTorch reimplementation of MBPO (When to trust your model: model-based policy optimization)

Dependency

Please refer to ./requirements.txt.

Usage

python ./mbpo/scripts/run_mbpo.py

# you can also overwrite hyperparams by passing args, e.g.
python ./mbpo/scripts/run_mbpo.py --set seed=0 verbose=1 device="'cuda:0'"

hyperparams in ./configs/mbpo.yaml

Credits

  1. vitchyr/rlkit
  2. facebookresearch/slbo