algorithms

Source Code / Acknowledgements

The algorithms in this folder are based on the following sources. Please refer to the works below:

• ACT: Code and Paper
• Diffusion Code and Paper
• IBC: Code and Paper
• GAIL: Code and Paper

Training the Policies

To train the policies, simply run:

conda activate <conda_env_name>
python imitate_<algorithm>.py

where:

• <algorithm> should be replaced with the method you intend to use: for example, diffusion
• <conda_env_name> is replaced according to the main README: for example, irl_torch

NOTE: you can pass --help to the script to view the available arguments.

When running the script, make sure to use the correct conda environment:

• conda activate irl_torch for: [ACT, Diffusion] (and experimental: BC)
• conda activate irl_tensorflow for: [IBC]
• conda activate irl_theano for: [GAIL, DAgger, BC]

NOTE: you can pass --help to the script to view the available arguments.