language_table_sim / README.md
mateoguaman's picture
Add dataset card
bbe871a verified
metadata
license: apache-2.0
task_categories:
  - robotics
tags:
  - LeRobot
configs:
  - config_name: default
    data_files: data/*/*.parquet

This dataset was created using LeRobot.

It is a LeRobot v3.0 conversion of the language_table_sim dataset from the Language Table project (Google Research).

The original RLDS data was streamed from gs://gresearch/robotics/language_table_sim/0.0.1/, converted to LeRobot v2.0 format preserving all fields (action, state, effector target, reward, done, instruction), then restructured to v3.0 with lossless video concatenation.

Dataset Description

Dataset Structure

meta/info.json:

{
    "codebase_version": "v3.0",
    "robot_type": "xarm",
    "total_episodes": 181020,
    "total_frames": 4484403,
    "total_tasks": 78627,
    "chunks_size": 1000,
    "fps": 10,
    "splits": {
        "train": "0:181020"
    },
    "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",
    "features": {
        "observation.images.rgb": {
            "dtype": "video",
            "shape": [
                360,
                640,
                3
            ],
            "names": [
                "height",
                "width",
                "channels"
            ],
            "video_info": {
                "video.fps": 10.0,
                "video.height": 360,
                "video.width": 640,
                "video.channels": 3,
                "video.codec": "libx264",
                "video.pix_fmt": "yuv420p",
                "video.is_depth_map": false,
                "has_audio": false
            }
        },
        "observation.state": {
            "dtype": "float32",
            "shape": [
                2
            ],
            "names": {
                "motors": [
                    "x",
                    "y"
                ]
            },
            "fps": 10
        },
        "observation.effector_target_translation": {
            "dtype": "float32",
            "shape": [
                2
            ],
            "names": {
                "motors": [
                    "x",
                    "y"
                ]
            },
            "fps": 10
        },
        "action": {
            "dtype": "float32",
            "shape": [
                2
            ],
            "names": {
                "motors": [
                    "x",
                    "y"
                ]
            },
            "fps": 10
        },
        "next.reward": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        },
        "next.done": {
            "dtype": "bool",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        },
        "timestamp": {
            "dtype": "float32",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        },
        "frame_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        },
        "episode_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        },
        "index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        },
        "task_index": {
            "dtype": "int64",
            "shape": [
                1
            ],
            "names": null,
            "fps": 10
        }
    },
    "data_files_size_in_mb": 100,
    "video_files_size_in_mb": 200
}