The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: ArrowInvalid
Message: Mismatching child array lengths
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/hdf5/hdf5.py", line 83, in _generate_tables
pa_table = _recursive_load_arrays(h5, self.info.features, start, end)
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/hdf5/hdf5.py", line 267, in _recursive_load_arrays
arr = _recursive_load_arrays(obj, features[path], start, end)
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/hdf5/hdf5.py", line 288, in _recursive_load_arrays
sarr = pa.StructArray.from_arrays(values, names=keys)
File "pyarrow/array.pxi", line 4304, in pyarrow.lib.StructArray.from_arrays
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: Mismatching child array lengthsNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
drawer_real60
60 real Franka demonstrations of a drawer task, in robomimic format. Baseline only: this file carries no object-pose labels. See What is missing before training an auxiliary-pose arm on it.
Each episode: open the drawer, pick up an arrow-shaped object from the table, place it inside. Two grasp cycles per demo (handle, then arrow), which is why every demo shows four gripper transitions.
Contents
| demos | 60 (mask/train 54, mask/valid 6) |
| samples | 30,374 |
| size | 0.81 GB |
| control rate | 15 Hz |
| session | drawer__09-13_22-55-44, collected 2026-09-13 |
env_args.env_name |
DrawerReal (provenance only -- there is no simulator) |
Per demo:
obs/agentview_image (T, 84, 84, 3) uint8 D435, scene
obs/robot0_eye_in_hand_image (T, 84, 84, 3) uint8 ZED Mini, wrist
obs/robot0_eef_pos (T, 3) float64 FK from joint_qpos, panda_hand_tcp
obs/robot0_eef_quat (T, 4) float64 xyzw
obs/robot0_gripper_qpos (T, 2) float64
actions (T, 7) float64 OSC_POSE deltas: dpos(3) | drot axis-angle(3) | grasp(1)
rewards, dones (T,) 1.0 / 1 on the final frame only
joint_qpos (T, 7) float64
Actions are derived from what the arm actually did (eef_pos[t+1] - eef_pos[t], and the
gripper-frame relative rotation), not from the teleop command stream. The final step repeats the
previous action because there is no t+1 for it.
Set experiment.rollout.enabled = false in the training config. There is no sim env behind these.
What is missing
obs/object and obs/aux_valid are absent, not zero-filled. Object-pose tracking for the
arrow had not been run when this was built, and writing zeros would have put something that reads
like a pose label into the file. An auxiliary-pose arm cannot train on this dataset; it will fail
at load time on the missing key, which is the intended behaviour.
A separate release will add the tracked arrow pose once FoundationPose has been run and validated.
Data quality
All 60 trajectories were checked before conversion:
- 60/60 open, keys and lengths consistent, 391-682 frames each
- depth valid on 92-95% of pixels, range 0.18-2.38 m
- no NaNs or jumps in the joint stream, no static trajectories
- D435: no duplicate frames anywhere in the session
- ZED Mini: 324 duplicated frames across 32 of the 60 demos (2-6% per demo, longest run
5 frames = 0.33 s). Intermittent USB bandwidth drops.
robot0_eye_in_hand_imagetherefore repeats a frame occasionally; the scene camera does not.
Related
- Downloads last month
- 33