This project is based on the open-source OmniGibson simulator and Behavior-1k benchmark. Follow their instructions to install. We suggest to install OmniGibson from source.
Clone this repo:
git clone https://github.com/keke-220/vlm-tamp.git
cd vlm-tamp
git submodule update --init --recursive
In the same conda environment as you run OmniGibson (conda activate omnigibson
), run our eval simply with:
python eval.py