Evaluation

Evaluation

Open-Loop Eval

export QWEN25_PATH=Qwen/Qwen2.5-VL-3B-Instruct
python scripts/open_loop_eval.py --model_path path_to_posttraining_ckpt --data_path path_to_validation_data --use_length 50
# If `--data_path` is omitted, the script defaults to the training dataset specified in the YAML config (`data.train_path`).

Note:
For inference, the model path (path_to_posttraining_ckpt, located in train.output_dir/checkpoints/*/hf_ckpt) must include:

• weights in .safetensors format
• config.json
• lingbotvla_cli.yaml

Robotwin

export QWEN25_PATH=path_to_Qwen2.5-VL-3B-Instruct
python -m deploy.lingbot_vla_policy \
 --model_path path_to_posttraining_ckpt \
 --use_compile \
 --use_length 50 \
 --port port

Real-Robot Deployment

export QWEN25_PATH=path_to_Qwen2.5-VL-3B-Instruct
python -m deploy.lingbot_vla_policy \
 --model_path path_to_posttraining_ckpt \
 --use_compile \
 --use_length 25
# You can set --num_denoising_step to 5 if you want to speed up the evaluation.

---

Install LIBERO

git clone https://github.com/Lifelong-Robot-Learning/LIBERO.git libero # (here)
cd libero
pip install -e .

cd experiment/libero/libero
pip install -r req.txt

If can not import xxx from libero.libero please add the libero (here) path to the PYTHONPATH variable.

The results will be save to /project_root/Libero

• release_ensemble/ stores the log files (This directory can be changed by --local_log_dir variable)
• rollouts stores the videos