Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: TypeError
Message: Couldn't cast array of type struct<candidate_mode: string, evaluated_candidates: int64, meets_clearance_buffer: bool, quality: struct<async_contact_frames: int64, async_min_distance_m: double, success: bool>, yield_event: struct<arm: string, start_step: int64, steps: int64>> to null
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, 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 129, 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 489, 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 2818, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, 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/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2303, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1852, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2143, in cast_array_to_feature
return array_cast(
array,
...<2 lines>...
allow_decimal_to_str=allow_decimal_to_str,
)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1854, in wrapper
return func(array, *args, **kwargs)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2007, in array_cast
raise TypeError(f"Couldn't cast array of type {_short_str(array.type)} to {_short_str(pa_type)}")
TypeError: Couldn't cast array of type struct<candidate_mode: string, evaluated_candidates: int64, meets_clearance_buffer: bool, quality: struct<async_contact_frames: int64, async_min_distance_m: double, success: bool>, yield_event: struct<arm: string, start_step: int64, steps: int64>> to nullNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
RoboTwin put_object_cabinet: 50 parallel episodes with dynamic main camera and calibrated wrist FOV
This is a real RoboTwin 2.0 put_object_cabinet dataset. It contains 50 native-success episodes generated from real expert trajectories and collision-aware retiming:
- episodes 0–24: left arm picks and places the object; right arm opens the drawer;
- episodes 25–49: right arm picks and places the object; left arm opens the drawer.
The main camera remains RoboTwin's native dynamic camera. Both wrist cameras use the same calibrated nfov_h60_d50_p10_fov80 mount and update from their gripper pose on every rendered frame. The original RoboTwin wrist-camera view remains available in the source code as the default original preset.
Camera preset
| Field | Value |
|---|---|
| preset | nfov_h60_d50_p10_fov80 |
| image size | 320 × 240 |
| vertical FOV | 80° |
| horizontal FOV | 96.4183429694° |
| calibration manifest SHA-256 | d92c5c42c1714a8bca0e91ebfd6db2fbbc1c6c25024865baea57af9e7512c579 |
The exact gripper-to-camera transform is stored in retime_manifest.json, every HDF5 episode, and validation_report.json.
Validation
- Native task success: 50/50.
- Role balance: 25 left-object / 25 right-object.
- Qwen2.5-VL-7B-Instruct stage proposals accepted after GT proprioception/state checking: 50/50.
- Collision-aware retime safety margin: 0.06 m.
- Minimum capsule distance during asynchronous execution: 0.060084965954579315 m.
- Asynchronous inter-arm contact frames: 0.
- Episodes containing reproducible yield/wait behavior: 6.
- Main-camera unique-frame ratio: 1.0 for every episode.
- Maximum wrist-camera mount-distance error: 7.036092580792186e-08 m.
- Renderer: ray tracing with OptiX denoiser for all 50 episodes.
validation_report.json contains dataset-wide metrics and per-episode camera, renderer, success, collision, and artifact-integrity receipts. Per-episode generation receipts are in validation/.
Contents
data/episodeN.hdf5: state, action, camera streams, calibration and retime metadata.video/episodeN.mp4: episode preview video.instructions/episodeN.json: task language annotations.validation/episodeN.json: native replay and validation receipt.metadata/annotations/episodeN.json: MLLM proposal plus GT repair/acceptance evidence.metadata/source_manifest.json: source episode, seed, role and source hashes.metadata/suite_manifest.json: full 50-episode generation-suite receipt.retime_manifest.json: retime policy and episode results.validation_report.json: independent final validation.SHA256SUMS: checksums for all published files except itself.
Provenance
- parallelVLA generation commit:
327417d7102b1fb2ba372e4a38413a7d52362a60. - parallelVLA final validator commit:
f3b8a310aa22adf5e80a31eb7de72ca0af6a29aa. - RoboTwin commit:
c3ddfa8b97d5519efa828b075999bd0006778e5e. - Stage model:
Qwen/Qwen2.5-VL-7B-Instruct. - Safety policy: collision-aware yield retime, 6 cm asynchronous safety margin, 4-frame synchronization hold.
Camera caveat
The active object arm briefly points its wrist camera into the cabinet during placement. The validator therefore observes up to 60 consecutive low-luminance frames and a maximum per-episode wrist dark-frame fraction of 0.285714. This is localized to the cabinet insertion phase; the camera mount stays within the transform tolerance above, and it is not the earlier failure mode where a camera was mounted inside a robot joint.
- Downloads last month
- 259