Skip to content

meta-RL soft actor-critic with BRUNO for task inference

Notifications You must be signed in to change notification settings

IraKorshunova/bruno-sac

Repository files navigation

BrunoSAC

Exchangeable Models in Meta Reinforcement Learning
I. Korshunova, J. Degrave, J. Dambre, A. Gretton, F. Huszár
Lifelong Learning Workshop at ICML 2020

Requirements

The code was used with the following settings:

  • python3
  • tensorflow-gpu==1.14.0
  • tensorflow-probability==0.7.0
  • gym==0.17.1
  • mujoco-py== 2.0.2.9
  • mujoco200

Training and testing

To train and then test BrunoSAC on Cheetah-Dir run:

python meta_cheetah_dir.py --train 
python meta_cheetah_dir.py --test

Similarly, for the oracle:

python meta_cheetah_dir.py --train --oracle
python meta_cheetah_dir.py --test --oracle

To plot the learning curves and test rewards:

python -m plots.plot_train_cheetah_dir 
python -m plots.plot_test_cheetah_dir

The same commands can be used with meta_cheetah_vel.py for the Cheetah-Vel experiments.

Questions?

Please send an email to [email protected], and I'll be happy to answer.

Releases

No releases published

Packages

No packages published

Languages