Hugging Face's logo

Datasets:
rchu233
/
kinder-shelf-sweep

Tasks:
Robotics
Tags:
kinder
mujoco
hdf5
world-model
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
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
task: string
env_id: string
seed: int64
num_episodes: int64
successes: int64
success_rate: double
max_steps: int64
open_loop_horizon: int64
texture_colorspace: string
action_mode: string
subgoal_source: string
skipped_episodes: list<item: null>
  child 0, item: null
mean_steps_success: double
success_rate_if_capped: struct<200: double, 300: double, 400: double, 500: double, 600: double, 700: double, 800: double, 10 (... 11 chars omitted)
  child 0, 200: double
  child 1, 300: double
  child 2, 400: double
  child 3, 500: double
  child 4, 600: double
  child 5, 700: double
  child 6, 800: double
  child 7, 1000: double
subtask_rates: struct<cube_lifted: double>
  child 0, cube_lifted: double
steps: int64
max_n_goal: int64
episode: int64
first_goal_step: null
wall_s: double
success: bool
mean_infer_s: double
final_n_goal: int64
cube_final_z: double
first_lift_step: int64
cube_lifted: bool
n_goal_total: int64
to
{'seed': Value('int64'), 'success': Value('bool'), 'steps': Value('int64'), 'mean_infer_s': Value('float64'), 'max_n_goal': Value('int64'), 'final_n_goal': Value('int64'), 'n_goal_total': Value('int64'), 'first_goal_step': Value('null'), 'cube_lifted': Value('bool'), 'first_lift_step': Value('int64'), 'cube_final_z': Value('float64'), 'episode': Value('int64'), 'wall_s': Value('float64')}
because column names don't match
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 544, 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 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, 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 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, 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 361, 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 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              task: string
              env_id: string
              seed: int64
              num_episodes: int64
              successes: int64
              success_rate: double
              max_steps: int64
              open_loop_horizon: int64
              texture_colorspace: string
              action_mode: string
              subgoal_source: string
              skipped_episodes: list<item: null>
                child 0, item: null
              mean_steps_success: double
              success_rate_if_capped: struct<200: double, 300: double, 400: double, 500: double, 600: double, 700: double, 800: double, 10 (... 11 chars omitted)
                child 0, 200: double
                child 1, 300: double
                child 2, 400: double
                child 3, 500: double
                child 4, 600: double
                child 5, 700: double
                child 6, 800: double
                child 7, 1000: double
              subtask_rates: struct<cube_lifted: double>
                child 0, cube_lifted: double
              steps: int64
              max_n_goal: int64
              episode: int64
              first_goal_step: null
              wall_s: double
              success: bool
              mean_infer_s: double
              final_n_goal: int64
              cube_final_z: double
              first_lift_step: int64
              cube_lifted: bool
              n_goal_total: int64
              to
              {'seed': Value('int64'), 'success': Value('bool'), 'steps': Value('int64'), 'mean_infer_s': Value('float64'), 'max_n_goal': Value('int64'), 'final_n_goal': Value('int64'), 'n_goal_total': Value('int64'), 'first_goal_step': Value('null'), 'cube_lifted': Value('bool'), 'first_lift_step': Value('int64'), 'cube_final_z': Value('float64'), 'episode': Value('int64'), 'wall_s': Value('float64')}
              because column names don't match

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.

KinDER Shelf3D + SweepIntoDrawer3D (adaptive-reasoning)

Data for world-model and VLA training on two KinDER tasks (Shelf3D-o1 and SweepIntoDrawer3D-o5). First built 2026-10-20; public and extended with the pi05 rollouts and labels on 2026-11-07. Code and notes: GitHub RobertCHU01/adaptive-reasoning (project_knowledge/, project_knowledge/team_setup.md).

The official human demos are not copied here. Download them from the public kinder-bench/kinder-datasets (shelf3d_106_new.hdf5, sweep3d_100.hdf5). Raw human .p demos are tracked in the kindergarden repo under demos/. The LeRobot versions used by OpenPi are in rchu233/kinder-lerobot.

Layout

Path Size Content
wap_kinder_ft_bg/ 0.8 GB World Action Planner finetune data: 4-view videos + pose (skeleton) videos. metadata.csv (239 rows, planner/human demos), metadata_vla.csv (+100 pi05 rollout rows)
visual_hdf5/Shelf3D-o{1,2,3}.hdf5 3.5 / 10.5 / 16.8 GB Bilevel-planner demos rendered for visual policies, 110 successful episodes each
raw/Shelf3D-o{1,2,3}/, raw/SweepIntoDrawer3D-o5/ ~40 MB Raw planner trajectories (<seed>/attempt_N.p) behind the HDF5 files
raw/pi05_rollouts/{shelf,sweep}/episode_XXX.p 45 MB 50 + 50 rollouts of pi05 pi05_kinder_official step 29999 (seed 301, job 4046519; Shelf 9/50, Sweep 14/50 succeed), raw-demo format
wm_test_rollouts/pi05_29999_seed401/ 15 MB Held-out pi05 rollouts (seed 401, 5 per task) for test_wm_ckpt.py
eef_actions/ 4 MB Per-demo end-effector targets (make_eef_actions.py), used to build kinder/shelf_sweep_subgoal
subgoal_labels/ 0.3 MB Rule-based subgoal + 3D point labels of the 206 human demos (label_subgoals.py)
vlm_labels/ 0.6 GB 10,272 (overview image, prompt, answer) examples for finetuning a VLM subgoal planner (export_vlm_labels.py)

wap_kinder_ft_bg/

Rendered with experiments/wap_kinder_ft/build_dataset.py (scene_bg=True). metadata.csv has 239 rows with a split column:

Task training validation holdout
Shelf3D-o1 100 2 17
SweepIntoDrawer3D-o5 100 2 18

metadata_vla.csv = metadata.csv + 100 vla30k_* training rows (pi05 rollouts, successes and failures; columns source, success, failure_mode, eval_result). The batch-64 WAP run (rchu233/wap-kinder-vla30k-b64) trained on it.

  • video_path, cond_path, prompt_embed_path, negative_prompt_embed_path are relative to wap_kinder_ft_bg/, so pass dataset.data_root=<download>/wap_kinder_ft_bg. The prompt embeddings are copies of WAP's robosuite_default_{,neg_}prompt.pt.
  • demo records provenance only: raw/... is relative to this repo's root, kindergarden/demos/... to the kindergarden repo.
  • Shelf training episodes come from raw/Shelf3D-o1 (planner). Sweep training uses human demos 3–102; demos 0–2 and the 9 human Shelf demos are the probe holdout.
  • Pose videos are yuv444p on purpose; yuv420p blurs the 1 px skeleton colours.

visual_hdf5/

Written by kindergarden/scripts/visual_dataset_queue.py. Every episode passed terminated=True, truncated=False.

File Episodes Steps
Shelf3D-o1.hdf5 110 11 240
Shelf3D-o2.hdf5 110 32 732
Shelf3D-o3.hdf5 110 52 501 
data/demo_N/actions              (T, 11) float32
data/demo_N/obs/robot_state      (T, 22) float32
data/demo_N/obs/env_state        (T, 23 for o1) float32
data/demo_N/obs/overview_image   (T, 224, 224, 3) uint8
data/demo_N/obs/base_image       (T, 224, 224, 3) uint8
data/demo_N/obs/wrist_image      (T, 224, 224, 3) uint8

Shelf3D-o3 was collected before the 2026-10-07 robot-spawn fix in kindergarden, so its episodes start from a random robot pose and replaying raw/Shelf3D-o3 with the fixed JSON will not reproduce them. o1 and o2 are unaffected.

Download 

hf download rchu233/kinder-shelf-sweep --repo-type dataset --local-dir <data_dir>/kinder-shelf-sweep
# only the WAP set:
hf download rchu233/kinder-shelf-sweep --repo-type dataset --include "wap_kinder_ft_bg/*" --local-dir <data_dir>/kinder-shelf-sweep

With hf CLI ≥ 1.0 give one --include per pattern.
Downloads last month
49
Total file size:
32.3 GB