Skip to content
/ BMIL Public

Pytorch code for "Learning Belief Representations for Imitation Learning in POMDPs" (UAI 2019)

Notifications You must be signed in to change notification settings

tgangwani/BMIL

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

This repo contains code for our paper Learning Belief Representations for Imitation Learning in POMDPs published at UAI 2019.

The code was tested with the following packages:

  • python 3.6.6
  • pytorch 0.4.1
  • gym 0.10.8

Running command

To run MuJoCo experiments, use the script run_mujoco.sh with the following usage:

bash run_mujoco.sh [env] [belief_loss_type] [belief_regularization]

BMIL results can be reproduced with bash run_mujoco.sh [env] task_aware True

Expert trajectories

Please update the path to expert trajectories in the file "code/conf/envParams.yaml". Also see the storage requirements in "code/expert_envs.py" and modify as per convenience.

Credits

The base for this code is provided by DVRL, which itself utilizes methods from this. We also use OpenAI baselines helpers for vectorized environments.

About

Pytorch code for "Learning Belief Representations for Imitation Learning in POMDPs" (UAI 2019)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published