Installation guide for RIPT-VLA

To run RIPT-VLA + QueST, you only need to install steps 0, 1, and 2.

To run RIPT-VLA + OpenVLA-OFT, you will additionally need to install steps 3.

0. Base RIPT-VLA environment

Create and activate conda environment

conda create -n ript_vla python=3.10.14
conda activate ript_vla

Install PyTorch

python -m pip install torch==2.2.0 torchvision==0.17.0

1. QueST

Install dependency (mostly inherited from QueST)

git clone https://github.com/Ariostgx/ript-vla.git
cd ript-vla
python -m pip install -e .

The codebase and dependency is based on the QueST.

2. LIBERO

Library

git clone https://github.com/Lifelong-Robot-Learning/LIBERO.git
cd LIBERO
python -m pip install -e .

Dataset

Please download the libero data following official docs.

3. OpenVLA-OFT

Create and activate a new conda environment for openvla-oft + RIPT-VLA

• We recommend creating a new conda environment for openvla-oft + RIPT-VLA to avoid dependency conflicts with QueST.

conda create -n ript_vla_openvla_oft python=3.10 -y
conda activate ript_vla_openvla_oft
pip3 install torch torchvision torchaudio

Clone openvla-oft repo and pip install to download dependencies

git clone https://github.com/moojink/openvla-oft.git
cd openvla-oft
pip install -e .

Install accleration libraries

#   =>> If you run into difficulty, try `pip cache remove flash_attn` first
pip install packaging ninja
ninja --version; echo $?  # Verify Ninja --> should return exit code "0"
pip install "flash-attn==2.7.4.post1" --no-build-isolation
pip install accelerate==1.6.0

Install RIPT-VLA dependency

cd ript-vla
python -m pip install -e .

The codebase and dependency is based on the QueST.

Install LIBERO

cd LIBERO
python -m pip install -e .

The codebase and dependency is based on the OpenVLA-OFT.