assests / common /source /starBenchserver /benchmarks /simBenchmarks /LIBERO /environment /verify_environment.py
| """Environment-only diagnostic; no evaluated model and no benchmark score.""" | |
| from __future__ import annotations | |
| import argparse | |
| import json | |
| from pathlib import Path | |
| from benchkit.artifacts import atomic_write_json | |
| from benchmarks.simBenchmarks.LIBERO.bench_server.adapter import FIRST_PASS_SEED, LiberoBenchmarkAdapter, task_catalog | |
| def main() -> None: | |
| parser = argparse.ArgumentParser(description=__doc__) | |
| parser.add_argument("--task-id", type=int, default=0) | |
| parser.add_argument("--suite", choices=tuple(task_catalog()), default="libero_10") | |
| parser.add_argument("--scope", choices=("development", "task_coverage"), default="development") | |
| parser.add_argument("--render-gpu-device-id", type=int, default=1) | |
| parser.add_argument("--output-dir", type=Path, required=True) | |
| args = parser.parse_args() | |
| output = args.output_dir.resolve() | |
| output.mkdir(parents=True, exist_ok=False) | |
| result = { | |
| "kind": "environment_diagnostic_not_model_evaluation", | |
| "task_id": args.task_id, "suite": args.suite, "scope": args.scope, | |
| "policy_calls": 0, "diagnostic_steps": 0, | |
| "native_success_after_diagnostic_step": None, | |
| "artifacts": {}, "status": "initializing", "error": None, | |
| } | |
| atomic_write_json(output / "diagnostic_status.json", {"status": "initializing"}) | |
| session = None | |
| try: | |
| adapter = LiberoBenchmarkAdapter(task_ids=[args.task_id], suite=args.suite, scope=args.scope, | |
| render_gpu_device_id=args.render_gpu_device_id) | |
| episode = adapter.episodes(FIRST_PASS_SEED)[0] | |
| session = adapter.open_episode(episode) | |
| session.bind_artifact_dir(output) | |
| observation = session.reset() | |
| transition = session.step((0.0,) * 6 + (-1.0,)) | |
| result["diagnostic_steps"] = 1 | |
| result["native_success_after_diagnostic_step"] = session.metrics()["success"] | |
| if observation.metadata or transition.observation.metadata or observation.state is not None: | |
| raise AssertionError("default RGB diagnostic leaked undeclared metadata/state") | |
| result["camera_shapes"] = {key: list(value.shape) for key, value in observation.images.items()} | |
| result["status"] = "completed" | |
| except Exception as exc: | |
| result["status"] = "failed" | |
| result["error"] = {"type": type(exc).__name__, "message": str(exc)} | |
| raise | |
| finally: | |
| try: | |
| if session is not None: | |
| try: | |
| result["artifacts"] = session.artifacts() | |
| finally: | |
| session.close() | |
| except Exception as exc: | |
| result["status"] = "failed" | |
| result["cleanup_error"] = {"type": type(exc).__name__, "message": str(exc)} | |
| raise | |
| finally: | |
| atomic_write_json(output / "diagnostic_result.json", result) | |
| atomic_write_json(output / "diagnostic_status.json", {"status": result["status"]}) | |
| print(json.dumps(result, indent=2, sort_keys=True)) | |
| if __name__ == "__main__": | |
| main() | |