{
  "schema_version": "rlinf.pi05-libero-eval-summary.v1",
  "model": {
    "name": "Pi05_few_shot_libero",
    "checkpoint": "global_step_16000",
    "step": 16000,
    "weights": "../actor/model_state_dict/full_weights.pt"
  },
  "status": "complete",
  "evaluation_parameters": {
    "benchmark": "LIBERO",
    "task_suites": [
      "libero_spatial",
      "libero_object",
      "libero_goal",
      "libero_10"
    ],
    "tasks_per_suite": 10,
    "episodes_per_task": 50,
    "trajectories_per_suite": 500,
    "evaluation_seed": 195,
    "environment_seed": 0,
    "reset_states": {
      "fixed": true,
      "ordered": true
    },
    "environment": {
      "auto_reset": true,
      "ignore_terminations": true,
      "reset_gripper_open": true,
      "reuse_environment_on_reset": true,
      "use_relative_reward": true
    },
    "observation": {
      "camera_height": 256,
      "camera_width": 256,
      "use_proprioception": true,
      "images_per_input": 2
    },
    "action": {
      "dimension": 7,
      "chunks_per_inference": 5,
      "adapter": "pi05_openpi"
    },
    "policy_inference": {
      "model_type": "openpi",
      "openpi_config_name": "pi05_libero",
      "checkpoint_format": "native",
      "denoising_steps": 10,
      "noise_level": 0.5,
      "noise_method": "flow_sde"
    },
    "suite_parameters": {
      "libero_spatial": {
        "max_episode_steps": 240,
        "reward_coefficient": 1.0
      },
      "libero_object": {
        "max_episode_steps": 280,
        "reward_coefficient": 1.0
      },
      "libero_goal": {
        "max_episode_steps": 320,
        "reward_coefficient": 1.0
      },
      "libero_10": {
        "max_episode_steps": 520,
        "reward_coefficient": 5.0
      }
    }
  },
  "coverage": {
    "expected_suites": 4,
    "completed_suites": 4,
    "missing_suites": [],
    "expected_tasks": 40,
    "completed_tasks": 40,
    "expected_trajectories": 2000,
    "completed_trajectories": 2000,
    "coverage_fraction": 1.0
  },
  "aggregate_metrics": {
    "scope": "all_expected_suites",
    "included_suites": [
      "libero_spatial",
      "libero_object",
      "libero_goal",
      "libero_10"
    ],
    "num_tasks": 40,
    "num_trajectories": 2000,
    "num_successes": 1446,
    "success_once": 0.723,
    "success_at_end": 0.14500000327825546,
    "mean_episode_length": 213.24049758911133,
    "mean_return": 1.2030000239610672,
    "mean_reward": 0.007830286282114685
  },
  "suite_results": [
    {
      "suite": "libero_spatial",
      "suite_label": "LIBERO-Spatial",
      "status": "complete",
      "task_result": "libero_spatial/task_result.json",
      "metrics": {
        "num_tasks": 10,
        "num_trajectories": 500,
        "num_successes": 381,
        "success_once": 0.762,
        "success_at_end": 0.15800000727176666,
        "mean_episode_length": 138.58999633789062,
        "mean_return": 0.7620000243186951,
        "mean_reward": 0.007456433493643999
      }
    },
    {
      "suite": "libero_object",
      "suite_label": "LIBERO-Object",
      "status": "complete",
      "task_result": "libero_object/task_result.json",
      "metrics": {
        "num_tasks": 10,
        "num_trajectories": 500,
        "num_successes": 461,
        "success_once": 0.922,
        "success_at_end": 0.15800000727176666,
        "mean_episode_length": 150.8179931640625,
        "mean_return": 0.921999990940094,
        "mean_reward": 0.006720130797475576
      }
    },
    {
      "suite": "libero_goal",
      "suite_label": "LIBERO-Goal",
      "status": "complete",
      "task_result": "libero_goal/task_result.json",
      "metrics": {
        "num_tasks": 10,
        "num_trajectories": 500,
        "num_successes": 364,
        "success_once": 0.728,
        "success_at_end": 0.1679999977350235,
        "mean_episode_length": 167.8679962158203,
        "mean_return": 0.7279999852180481,
        "mean_reward": 0.007377552334219217
      }
    },
    {
      "suite": "libero_10",
      "suite_label": "LIBERO-10",
      "status": "complete",
      "task_result": "libero_10/task_result.json",
      "metrics": {
        "num_tasks": 10,
        "num_trajectories": 500,
        "num_successes": 240,
        "success_once": 0.48,
        "success_at_end": 0.09600000083446503,
        "mean_episode_length": 395.6860046386719,
        "mean_return": 2.4000000953674316,
        "mean_reward": 0.009767028503119946
      }
    }
  ]
}