Quick Tour#

This tutorial walks through your first embodied evaluation in about 5 minutes using LIBERO Spatial + OpenPI π₀.₅. The example config is evaluations/libero/libero_spatial_openpi_pi05_eval.yaml.

Step 1: Install the Environment#

From the repository root:

bash requirements/install.sh embodied --model openpi --env libero
source .venv/bin/activate

With --env libero, the installer clones LIBERO into .venv/libero (or reuses an existing checkout when LIBERO_PATH is set) and appends it to PYTHONPATH in .venv/bin/activate. No separate LIBERO_PATH setup is required.

Step 2: Prepare the Model#

Download the SFT checkpoint from Hugging Face (RLinf/RLinf-Pi05-LIBERO-SFT) to a local directory:

huggingface-cli download RLinf/RLinf-Pi05-LIBERO-SFT --local-dir ./RLinf-Pi05-LIBERO-SFT

Model hub: RLinf/RLinf-Pi05-LIBERO-SFT. You can also override rollout.model.model_path on the command line at launch time.

Step 3: Launch Evaluation#

bash evaluations/run_eval.sh libero libero_spatial_openpi_pi05_eval \
  rollout.model.model_path=./RLinf-Pi05-LIBERO-SFT

When the config name starts with libero_, you can omit the benchmark argument:

bash evaluations/run_eval.sh libero_spatial_openpi_pi05_eval \
  rollout.model.model_path=./RLinf-Pi05-LIBERO-SFT

Step 4: Check Results#

  • The terminal prints metrics such as eval/success_once and eval/return

  • Log directory: logs/<timestamp>-libero_spatial_openpi_pi05_eval/eval_embodiment.log

  • When env.eval.video_cfg.save_video: True, videos are saved under <log_path>/video/eval/

See Logs and Results for more details.

Next Steps#