OakInk2-SimEnv-IsaacGym

1. Setup asset files.

   Download tarballs from huggingface. Extract the tarball under asset/.

   For evaluation, download [obj_interior.tar.xz]() and pour_eval.tar.xz, and extract them under asset/oi2_dev.

   The directory structure:

   asset
   |--- mano_mod
   |--- oi2_dev
   |   |--- obj
   |   |--- obj_raw
   |   |--- obj_interior
   |   |--- pour
   |   |   |--- demo_raw
   |   |   |--- meta
   |   |   |--- ref_raw
   |   |   `--- task_list_rh__train.json
   |   `--- pour_eval
   |       |--- meta
   |       |--- ref_raw
   |       `--- task_list_rh__test.json
   `--- shadowhand_no_forearm

2. Setup the enviroment.

   1. Create a virtual env of python 3.8. This can be done by either conda or python package venv.

      1. conda approach

         conda create -p ./.conda python=3.8
         conda activate ./.conda

      2. venv approach First use pyenv or other tools to install a python intepreter of version 3.8. Here 3.8.19 is used as example:

         pyenv install 3.8.19
         pyenv shell 3.8.19

         Then create a virtual environment:

         python -m venv .venv --prompt OakInk2-SimEnv-IsaacGym
         . .venv/bin/activate

   2. Install the dependencies

      Make sure all bundled dependencies are there.

      git submodule update --init --recursive --progress

      Download Issac Gym Preview 4. Extract it under thirdparty/isaac_gym/. The directory structure:

      thirdparty
      |--- isaac_gym
      |   `--- isaacgym
      |--- ...

      Install the pip requirements.

      # essential
      pip install -r req_py38_isaacgym.example.txt
      # essential with devtools
      pip install -r req_py38_isaacgym_dev.example.txt

3. View the example script.

   python -m script.test_env_pour

   If you want to use reinforcement learning algorithms in this toolkit, there is an example in src/dyn_mf/env/oi2_dev.py to set the reward and observation functions. The observation and reward functions set in the current example file are both blank placeholders to be filled.