lambdavi/libero_unified_nvidia

LIBERO Unified (NVIDIA LeRobot v3)

Single LeRobot v3 dataset merging all four NVIDIA LIBERO_LeRobot_v3 suites with global language-keyed task indices (0–39). It is useful for non-language multi-task imitation learning models.

Source

Suite	Role
libero_10	Long-horizon (10 tasks)
libero_goal	Goal suite (10 tasks)
libero_object	Object suite (10 tasks)
libero_spatial	Spatial suite (10 tasks)

Built from nvidia/LIBERO_LeRobot_v3 (local merge; videos copied, not re-encoded).

What we changed

1. Merged the four suite folders into one LeRobot root (meta/, data/, videos/).
2. Remapped task_index from each suite’s local 0–9 to a global 0–39 table keyed by the natural-language instruction (same order as HuggingFaceVLA/libero).
3. Kept 20 FPS, cameras observation.images.image + observation.images.wrist_image, robot franka.

Without the global remap, suite-local ids collide (e.g. spatial task 0 ≠ object task 0).

Task index layout

Range	Suite
0–9	libero_10
10–19	libero_goal
20–29	libero_object
30–39	libero_spatial

Eval: map (suite_name, env.task_id) → sim language → same global id (see gap_policy/libero_lerobot_adapter/task_index_map.py).

Stats

• Episodes: 1693
• Frames: 273465
• Tasks: 40
• FPS: 20

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "fps": 20,
    "features": {
        "observation.images.wrist_image": {
            "dtype": "video",
            "shape": [
                256,
                256,
                3
            ],
            "names": [
                "height",
                "width",
                "rgb"
            ],
            "info": {
                "video.height": 256,
                "video.width": 256,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.fps": 20,
                "video.channels": 3,
                "has_audio": false,
                "video.video_backend": "pyav",
                "video.g": null,
                "video.fast_decode": null,
                "video.preset": null,
                "video.crf": null,
                "video.extra_options": null,
                "is_depth_map": false
            }
        },
        "observation.images.image": {
            "dtype": "video",
            "shape": [
                256,
                256,
                3
            ],
            "names": [
                "height",
                "width",
                "rgb"
            ],
            "info": {
                "video.height": 256,
                "video.width": 256,
                "video.codec": "av1",
                "video.pix_fmt": "yuv420p",
                "video.fps": 20,
                "video.channels": 3,
                "has_audio": false,
                "video.video_backend": "pyav",
                "video.g": null,
                "video.fast_decode": null,
                "video.preset": null,
                "video.crf": null,
                "video.extra_options": null,
                "is_depth_map": false
            }
        },
        "observation.state": {
            "dtype": "float32",
            "shape": [
                8
            ],
            "names": {
                "motors": [
                    "x",
                    "y",
                    "z",
                    "axis_angle1",
                    "axis_angle2",
                    "axis_angle3",
                    "gripper",
                    "gripper"
                ]
            }
        },
        "action": {
            "dtype": "float32",
            "shape": [
                7
            ],
            "names": {
                "motors": [
                    "x",
                    "y",
                    "z",
                    "axis_angle1",
                    "axis_angle2",
                    "axis_angle3",
                    "gripper"
                ]
            }
        },
        "timestamp": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null
        },
        "frame_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "episode_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        },
        "task_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null
        }
    },
    "total_episodes": 1693,
    "total_frames": 273465,
    "total_tasks": 40,
    "chunks_size": 1000,
    "data_files_size_in_mb": 100,
    "video_files_size_in_mb": 200,
    "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
    "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
    "robot_type": "franka",
    "splits": {
        "train": "0:1693"
    }
}

Citation

BibTeX:

[More Information Needed]