# Create and activate conda environment
conda create -n openvla python=3.10 -y
conda activate openvla

# Install PyTorch. Below is a sample command to do this, but you should check the following link
# to find installation instructions that are specific to your compute platform:
# https://pytorch.org/get-started/locally/
conda install pytorch torchvision torchaudio pytorch-cuda=12.4 -c pytorch -c nvidia -y  # UPDATE ME!

# Clone and install the openvla repo
git clone https://github.com/openvla/openvla.git
cd openvla
pip install -e .

# Install Flash Attention 2 for training (https://github.com/Dao-AILab/flash-attention)
#   =>> 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==2.5.5" --no-build-isolation

基本没什么问题，中间numpy版本冲突了一下，换了一下opencv的版本

有报错或许可以在pip最后加上以下来解决

--use-pep517

LIBERO 模拟基准评估

可以安装在openvla目录下

git clone https://github.com/Lifelong-Robot-Learning/LIBERO.git
cd LIBERO
pip install -e .
此外，安装其他必需的软件包：

cd openvla
pip install -r experiments/robot/libero/libero_requirements.txt

这里安装libero，就算标黄也别用以下了，后面可能测试评估时会报错

--use-pep517

我们通过 LoRA (r=32) 在四个 LIBERO 任务套件上分别对 OpenVLA 进行了微调：LIBERO-Spatial、LIBERO-Object、LIBERO-Goal 和 LIBERO-10（也称为 LIBERO-Long）。这四个检查点在 Hugging Face 上可用：

其中一个，可以安装在openvla目录下

https://huggingface.co/openvla/openvla-7b-finetuned-libero-spatial

git clone https://huggingface.co/openvla/openvla-7b-finetuned-libero-spatial

回到openvla目录

python experiments/robot/libero/run_libero_eval.py \
  --model_family openvla \
  --pretrained_checkpoint openvla/openvla-7b-finetuned-libero-spatial \
  --task_suite_name libero_spatial \
  --center_crop True

应该可以开始跑了

如果遇到

    import libero
ModuleNotFoundError: No module named 'libero'

就是前面安装用了下面的，别用。再装一遍就行

--use-pep517

结果在rollouts里

 

50系显卡需要注意pytorch版本问题，注意下载2.6以上的

1.  2.6以上需要修改此处：找到__init__.py中torch.load,加上weights_only=False

init_states = torch.load(init_states_path, weights_only=False)

2.  AttributeError: 'MjRenderContextOffscreen' object has no attribute 'con'

可通过以下解决

export MUJOCO_GL=glx