Code for DPAIL: Training Diffusion Policy for Adversarial Imitation Learning without Policy Optimization (Accepted in NeurIPS 2025, link)
To use this code, you need the following versions of Python and PyTorch: Python >= 3.10 PyTorch >= 2.0.1 (install from the official website)
The Required Library
mujoco-py==2.1.2.14
gym==0.23.0
einops
python train_imitation.py --buffer buffers/Ant-v3/size10000 --cuda