We test the pretrained visual reprensetations in the few-shot visual imitation learning setting. The experiment design for this is based on the PVR paper.
To download the demonstration datasets, create a directory for the dataset by executing the following command:
mkdir -p benchmarks/mujoco_vc/visual_imitation/data/datasets
cd benchmarks/mujoco_vc/visual_imitation/data/datasets
Then, download the dataset with the following commands:
wget https://dl.fbaipublicfiles.com/eai-vc/mujoco_vil_datasets/metaworld-expert-v1.0.zip
unzip metaworld-expert-v1.0.zip
rm metaworld-expert-v1.0.zip
To run experiments, navigate to the visual_imitation
subdirectory, which contains launch scripts and config files, by executing the following commands:
cd stable-control-representations/benchmarks/mujoco_vc/visual_imitation/
To run the job, execute the following script:
./run_script.sh