my implementation with adversarial inverse reinforcement learning on gym(breakout and pendulum)
tensorflow 2.o
gym 0.14.0
cudatoolkit 10.0.130
cudnn 7.6.5
GPU driver version 435.21
baseline 0.1.5(branch tf2)
after you run generate_expert_data.py , Then you can run airl.sh to do this task.