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: CastError
Message: Couldn't cast
format: string
format_version: int64
tasks: list<item: struct<task: string, task_slug: string, source_file: string, manifest: string>>
child 0, item: struct<task: string, task_slug: string, source_file: string, manifest: string>
child 0, task: string
child 1, task_slug: string
child 2, source_file: string
child 3, manifest: string
created_at_utc: string
created_by: struct<script: string>
child 0, script: string
combined_from: list<item: struct<input_index: int64, root: string>>
child 0, item: struct<input_index: int64, root: string>
child 0, input_index: int64
child 1, root: string
num_input_roots: int64
num_tasks: int64
num_demos: int64
source_desc: string
source_type: string
commit_info: null
env_info: struct<env_id: string, env_kwargs: struct<obs_mode: string, control_mode: string, render_mode: strin (... 203 chars omitted)
child 0, env_id: string
child 1, env_kwargs: struct<obs_mode: string, control_mode: string, render_mode: string, reward_mode: string, sensor_conf (... 141 chars omitted)
child 0, obs_mode: string
child 1, control_mode: string
child 2, render_mode: string
child 3, reward_mode: string
child 4, sensor_configs: struct<shader_pack: string, width: int64, height: int64>
child 0, shader_pack: string
child 1, width: int64
child 2, height: int64
child 5, human_render_camera_configs: struct<shader_pack: string>
child 0, shader_pack: string
child 6, sim_backend: string
child 2, max_episode_steps: int64
episodes: list<item: struct<episode_id: int64, episode_seed: int64, control_mode: string, elapsed_steps: int64 (... 413 chars omitted)
child 0, item: struct<episode_id: int64, episode_seed: int64, control_mode: string, elapsed_steps: int64, reset_kwa (... 401 chars omitted)
child 0, episode_id: int64
child 1, episode_seed: int64
child 2, control_mode: string
child 3, elapsed_steps: int64
child 4, reset_kwargs: struct<options: null, seed: int64>
child 0, options: null
child 1, seed: int64
child 5, success: bool
child 6, strategy: struct<seed: int64, grasp_tilt_deg: double, grasp_offset: double, grasp_flip: bool, standoff: double (... 234 chars omitted)
child 0, seed: int64
child 1, grasp_tilt_deg: double
child 2, grasp_offset: double
child 3, grasp_flip: bool
child 4, standoff: double
child 5, lift: bool
child 6, lift_height: double
child 7, lift_dx: double
child 8, lift_dy: double
child 9, speed_transit: double
child 10, speed_insert: double
child 11, approach_dy: double
child 12, approach_dz: double
child 13, approach_pitch_deg: double
child 14, approach_yaw_deg: double
child 15, transit_planner: string
to
{'env_info': {'env_id': Value('string'), 'env_kwargs': {'obs_mode': Value('string'), 'control_mode': Value('string'), 'render_mode': Value('string'), 'reward_mode': Value('string'), 'sensor_configs': {'shader_pack': Value('string'), 'width': Value('int64'), 'height': Value('int64')}, 'human_render_camera_configs': {'shader_pack': Value('string')}, 'sim_backend': Value('string')}, 'max_episode_steps': Value('int64')}, 'commit_info': Value('null'), 'episodes': List({'episode_id': Value('int64'), 'episode_seed': Value('int64'), 'control_mode': Value('string'), 'elapsed_steps': Value('int64'), 'reset_kwargs': {'options': Value('null'), 'seed': Value('int64')}, 'success': Value('bool'), 'strategy': {'seed': Value('int64'), 'grasp_tilt_deg': Value('float64'), 'grasp_offset': Value('float64'), 'grasp_flip': Value('bool'), 'standoff': Value('float64'), 'lift': Value('bool'), 'lift_height': Value('float64'), 'lift_dx': Value('float64'), 'lift_dy': Value('float64'), 'speed_transit': Value('float64'), 'speed_insert': Value('float64'), 'approach_dy': Value('float64'), 'approach_dz': Value('float64'), 'approach_pitch_deg': Value('float64'), 'approach_yaw_deg': Value('float64'), 'transit_planner': Value('string')}}), 'source_type': Value('string'), 'source_desc': Value('string')}
because column names don't match
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 2951, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, 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 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, 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 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
format: string
format_version: int64
tasks: list<item: struct<task: string, task_slug: string, source_file: string, manifest: string>>
child 0, item: struct<task: string, task_slug: string, source_file: string, manifest: string>
child 0, task: string
child 1, task_slug: string
child 2, source_file: string
child 3, manifest: string
created_at_utc: string
created_by: struct<script: string>
child 0, script: string
combined_from: list<item: struct<input_index: int64, root: string>>
child 0, item: struct<input_index: int64, root: string>
child 0, input_index: int64
child 1, root: string
num_input_roots: int64
num_tasks: int64
num_demos: int64
source_desc: string
source_type: string
commit_info: null
env_info: struct<env_id: string, env_kwargs: struct<obs_mode: string, control_mode: string, render_mode: strin (... 203 chars omitted)
child 0, env_id: string
child 1, env_kwargs: struct<obs_mode: string, control_mode: string, render_mode: string, reward_mode: string, sensor_conf (... 141 chars omitted)
child 0, obs_mode: string
child 1, control_mode: string
child 2, render_mode: string
child 3, reward_mode: string
child 4, sensor_configs: struct<shader_pack: string, width: int64, height: int64>
child 0, shader_pack: string
child 1, width: int64
child 2, height: int64
child 5, human_render_camera_configs: struct<shader_pack: string>
child 0, shader_pack: string
child 6, sim_backend: string
child 2, max_episode_steps: int64
episodes: list<item: struct<episode_id: int64, episode_seed: int64, control_mode: string, elapsed_steps: int64 (... 413 chars omitted)
child 0, item: struct<episode_id: int64, episode_seed: int64, control_mode: string, elapsed_steps: int64, reset_kwa (... 401 chars omitted)
child 0, episode_id: int64
child 1, episode_seed: int64
child 2, control_mode: string
child 3, elapsed_steps: int64
child 4, reset_kwargs: struct<options: null, seed: int64>
child 0, options: null
child 1, seed: int64
child 5, success: bool
child 6, strategy: struct<seed: int64, grasp_tilt_deg: double, grasp_offset: double, grasp_flip: bool, standoff: double (... 234 chars omitted)
child 0, seed: int64
child 1, grasp_tilt_deg: double
child 2, grasp_offset: double
child 3, grasp_flip: bool
child 4, standoff: double
child 5, lift: bool
child 6, lift_height: double
child 7, lift_dx: double
child 8, lift_dy: double
child 9, speed_transit: double
child 10, speed_insert: double
child 11, approach_dy: double
child 12, approach_dz: double
child 13, approach_pitch_deg: double
child 14, approach_yaw_deg: double
child 15, transit_planner: string
to
{'env_info': {'env_id': Value('string'), 'env_kwargs': {'obs_mode': Value('string'), 'control_mode': Value('string'), 'render_mode': Value('string'), 'reward_mode': Value('string'), 'sensor_configs': {'shader_pack': Value('string'), 'width': Value('int64'), 'height': Value('int64')}, 'human_render_camera_configs': {'shader_pack': Value('string')}, 'sim_backend': Value('string')}, 'max_episode_steps': Value('int64')}, 'commit_info': Value('null'), 'episodes': List({'episode_id': Value('int64'), 'episode_seed': Value('int64'), 'control_mode': Value('string'), 'elapsed_steps': Value('int64'), 'reset_kwargs': {'options': Value('null'), 'seed': Value('int64')}, 'success': Value('bool'), 'strategy': {'seed': Value('int64'), 'grasp_tilt_deg': Value('float64'), 'grasp_offset': Value('float64'), 'grasp_flip': Value('bool'), 'standoff': Value('float64'), 'lift': Value('bool'), 'lift_height': Value('float64'), 'lift_dx': Value('float64'), 'lift_dy': Value('float64'), 'speed_transit': Value('float64'), 'speed_insert': Value('float64'), 'approach_dy': Value('float64'), 'approach_dz': Value('float64'), 'approach_pitch_deg': Value('float64'), 'approach_yaw_deg': Value('float64'), 'transit_planner': Value('string')}}), 'source_type': Value('string'), 'source_desc': Value('string')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
PlugCharger 4 mm: 600 motion-planning demos
600 successful demonstrations of the ManiSkill PlugCharger task with a 4 mm hole clearance, recorded with three task cameras plus ManiSkill's base camera. Every demo was produced by a motion-planning expert; no human teleoperation or learned policy was involved.
The data comes in two forms:
| Path | Contents |
|---|---|
raw/ |
The original ManiSkill trajectory files (.h5 plus .json metadata), five files totalling 600 demos |
small_files/sd0_plug_charger_hole4mm_sidecam_fixedgrasp_600mp_combined/ |
All 600 demos converted to the robopolicy small-files format: full resolution, one demo per shard, with normalization statistics |
Raw files
| File | Demos | Scene seeds | Strategy seed |
|---|---|---|---|
sd0_plug_charger_hole4mm_sidecam_fixedgrasp_100 |
100 | 0 to 99 | 0 |
sd0_plug_charger_hole4mm_sidecam_fixedgrasp_extra200 |
200 | 2000 to 2201 | 1 |
sd0_plug_charger_hole4mm_sidecam_fixedgrasp_extra300_part1 |
100 | 10000 to 10101 | 3 |
sd0_plug_charger_hole4mm_sidecam_fixedgrasp_extra300_part2 |
100 | 11000 to 11099 | 4 |
sd0_plug_charger_hole4mm_sidecam_fixedgrasp_extra300_part3 |
100 | 12000 to 12099 | 5 |
Scene seeds are the values passed to env.reset(seed=...). A few seeds in a range are missing where the planner failed; only successful demos are stored. The strategy seed drives the planner's own randomization and is independent of the scene seed. Each episode's sampled strategy is recorded in the .json file under strategy.
Environment
- Task:
PlugCharger-Hole4mm-SideCam-v1, a ManiSkill 3 PlugCharger variant with a 4 mm hole clearance and an added side camera on the socket. - Robot and control: Franka Panda,
control_mode=pd_joint_pos. Actions are 8-dimensional: 7 joint position targets plus the gripper command. - Observations:
obs_mode=rgbd, every camera at 256 x 256. The cameras arehand_camera,third_person_camera,side_cameraandbase_camera. - Episode limit: 300 steps. A demo that only succeeded after step 300 was rejected rather than stored.
Expert
A strategy-diverse motion planner, configured as follows:
- Grasp: fixed at the stock pose, 15 degree tilt and zero offset.
- Diversity: sampled per episode for the transit to the socket, the approach to the socket, and the motion speed.
- Ending: each episode finishes with a 30-step closed-gripper pause at the insert pose, instead of releasing the charger.
Converted dataset
The small_files folder holds the five raw files converted one demo per shard at full resolution, with no image downsampling, and then combined into one dataset. Camera frames are stored as one MP4 per camera per shard. The folder contains:
plug_charger/shards/: one HDF5 file per demo with its state and action data.plug_charger/videos_shards/<camera>/: one MP4 per demo, each with a JSON file of frame metadata.plug_charger/demo_manifest.json: the demo index.dataset_manifest.json: which raw files were combined, and when.normalization_stats.npz: the statistics the trainer uses.
- Downloads last month
- 190