Datasets:

rooty2020
/

trellis_500k_subtask

Tasks:

Robotics

Modalities:

Tags:

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 80, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

trellis_500k_subtask — DROID subtask split (3D point tracks + per-subtask actions)

Per-subtask temporal split of the DROID 3D point-track episodes. Every episode is losslessly partitioned into its RoboInter subtasks; each subtask ships its own sliced action sequence alongside the 3D point tracks, video and calibration.

Episodes: 28,308 Subtasks: 89,312
Subtasks with actions: 89,312
Episodes with actions: 28,308 / 28,308 (without: 0)
Shards: 12 tars subtask_split-000.tar … (~335 GB total)
fps: 10

Layout

<episode_id>/
  subtask_00/  view1.mp4 view2.mp4  robot.npz gripper.npz  meta.json  action.parquet
  subtask_01/  ...
  ...

Each subtask_XX/ is a self-contained sample:

file	contents
`view1.mp4`, `view2.mp4`	the two DROID exterior views, sliced to this subtask
`robot.npz`	3D robot-arm point tracks `(T,512,3)` in the view1 camera frame
`gripper.npz`	3D gripper point tracks `(T,256,3)`
`meta.json`	calibration (per-view `T_cam_world`, intrinsics), subtask instruction, `n_frames`
`action.parquet`	RoboInter actions for exactly this subtask's frames

Matching actions to the point tracks

The tracks and the actions are the same T frames at 10 fps, in order — row t of action.parquet corresponds to frame t of robot.npz/gripper.npz/the videos in that subtask. Nothing to re-index.

action.parquet is the episode's RoboInter parquet sliced with the manifest's half-open frame_range [start, end) (so len == n_frames). Columns:

action (7,), state (7,), frame_index, timestamp, episode_name
other_information.* — joint position/velocity, cartesian & gripper velocity, TCP/wrist pose and deltas, gripper open state, is_first/last/terminal, extra language.

Reconstruct the full-episode action stream by concatenating a subtask's rows in order:

import pandas as pd, glob
dfs = [pd.read_parquet(f"{ep}/subtask_{i:02d}/action.parquet") for i in range(n)]
episode_actions = pd.concat(dfs, ignore_index=True)   # == original RoboInter episode

manifest.json lists every episode's per-subtask frame_range, primitive_skill and subtask_text, plus each shard's sha256/size — inspect without untarring.

Coverage: all 28,308 episodes have actions. For 44 episodes RoboInter supplies 1–2 fewer trailing frames than the media, so only those episodes' final subtask action.parquet has 1–2 rows fewer than its n_frames (every earlier subtask is exact). They are listed in manifest.json → tail_short_episodes.

Provenance & license

3D point tracks / subtask segmentation: DROID (CC-BY-4.0) + subtask annotations from InternRobotics/RoboInter-Data (Annotation_pure).
Actions: InternRobotics/RoboInter-Data (Annotation_with_action_lerobotv21/lerobot_droid_anno).
RoboInter data is CC-BY-NC-SA-4.0 — this release inherits it (non-commercial, share-alike).

Downloads last month: 46

Total file size:

338 GB