Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code: DatasetGenerationError
Exception: CastError
Message: Couldn't cast
episode_0: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
episode_1: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 62 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
episode_2: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
chil
...
ure_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
episode_498: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
episode_499: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
unseen: list<item: string>
child 0, item: string
seed: int64
seen: list<item: string>
child 0, item: string
to
{'seed': Value('int64'), 'seen': List(Value('string')), 'unseen': List(Value('string'))}
because column names don't match
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
for key, table in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
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
episode_0: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
episode_1: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 62 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
episode_2: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
chil
...
ure_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
episode_498: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
episode_499: struct<cluttered_table_info: list<item: null>, texture_info: struct<wall_texture: null, table_textur (... 75 chars omitted)
child 0, cluttered_table_info: list<item: null>
child 0, item: null
child 1, texture_info: struct<wall_texture: null, table_texture: null>
child 0, wall_texture: null
child 1, table_texture: null
child 2, info: struct<{A}: string, {B}: string, {a}: string, {C}: string>
child 0, {A}: string
child 1, {B}: string
child 2, {a}: string
child 3, {C}: string
unseen: list<item: string>
child 0, item: string
seed: int64
seen: list<item: string>
child 0, item: string
to
{'seed': Value('int64'), 'seen': List(Value('string')), 'unseen': List(Value('string'))}
because column names don't match
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/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1880, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
seed int64 | seen list | unseen list |
|---|---|---|
0 | [
"Ensure the brown bread loaf and the bread with rough texture are both inside the medium-sized rectangular breadbasket.",
"Pick up both the golden crusted bread slice and the light golden brown bread for the brown breadbasket.",
"Transfer both the soft golden baked bread and the golden bread loaf to the light b... | [
"Move both the bread loaf medium-sized and the rectangular bread loaf inside the rectangular light brown breadbasket.",
"Put the two breads the small loaf-shaped bread and the bread loaf with soft surface in the medium light brown breadbasket.",
"Transfer the bread loaf medium-sized and the bread with golden cr... |
1 | [
"Place the soft golden baked bread in the light brown wicker breadbasket.",
"Pick the square loaf of bread using the right arm and set it in the brown breadbasket.",
"Move the rectangular bread soft inside from the table into the breadbasket with rounded edges.",
"Take the golden bread and drop it in the brow... | [
"Pick up the soft rectangular bread loaf and set it in the rectangular light brown breadbasket.",
"Transfer the bread the soft rectangular bread loaf to the rectangular breadbasket.",
"Use the right arm to move the small loaf-shaped bread into the rectangular light brown breadbasket.",
"Put the bread loaf med... |
2 | [
"Put the soft golden bread and the light golden brown bread inside the breadbasket made of woven wood.",
"Lift the golden bread loaf with soft inside and the bread loaf with bumpy surface and place them in the palm-sized breadbasket.",
"Ensure the smooth crusty small bread and the bread with rough texture are b... | [
"Collect the round braided loaf and the bread with golden crust inside the beige breadbasket.",
"Place the round braided loaf and the bread with golden crust into the beige breadbasket.",
"Put the two breads the palm-sized braided golden bread and the rectangular bread loaf in the light wooden breadbasket.",
... |
3 | [
"Collect the bread with rough texture and the golden brown bread in the woven light beige breadbasket.",
"Place the bread loaf with uneven edges and the golden brown bread in the beige breadbasket.",
"Transfer both the hand-sized bread block and the braided bread loaf to the breadbasket with textured sides.",
... | [
"Put the two breads the bread with golden crust and the round braided loaf in the breadbasket with woven pattern.",
"Transfer the bread with golden crust and the round braided loaf from the table to the shallow rectangular breadbasket.",
"Collect the rectangular bread loaf and the round braided loaf inside the ... |
4 | [
"Pick up both the small bread and the bread with golden surface color for the light brown breadbasket.",
"Collect the soft doughy bread and the golden brown loaf-shaped bread in the smooth-textured breadbasket.",
"Put the round bread with crusty texture and the golden crusted bread slice inside the oval breadba... | [
"Collect the round braided loaf and the small loaf-shaped bread inside the medium-sized breadbasket.",
"Move both the palm-sized braided golden bread and the small loaf-shaped bread inside the beige breadbasket.",
"Place the palm-sized braided golden bread and the bread loaf medium-sized into the breadbasket wi... |
5 | [
"Place the soft bread and the small golden baked bread in the smooth textured brown breadbasket.",
"Transfer both the cuboid bread and the soft doughy bread to the woven rectangular breadbasket.",
"Deposit the square-shaped bread and the round bread with braided top inside the breadbasket with smooth woven text... | [
"Transfer the bread loaf with soft surface and the braided top round bread from the table to the rectangular light brown breadbasket.",
"Put the two breads the bread loaf with soft surface and the round braided loaf in the rectangular breadbasket.",
"Collect the rectangular bread loaf and the palm-sized braided... |
6 | [
"Place the bread with rounded corners and the rectangular bread soft inside in the medium breadbasket.",
"Grab the brown bread loaf and the rectangular bread soft inside and put them in the breadbasket with flat base.",
"Transfer both the slice-shaped bread and the golden crusted bread slice to the breadbasket ... | [
"Transfer the soft rectangular bread loaf and the bread loaf medium-sized from the table to the round beige breadbasket.",
"Collect the bread loaf medium-sized and the bread loaf medium-sized inside the beige breadbasket.",
"Move both the bread loaf medium-sized and the small loaf-shaped bread inside the beige ... |
7 | [
"Lift the bread with rounded edges and the bread with soft texture and place them in the oval-shaped breadbasket.",
"Relocate the golden brown bread and the square-shaped bread piece into the basket the beige basket with white grid design.",
"Place both breads the golden baked bread and the sponge-textured gold... | [
"Transfer the square bread with light spots and the square bread with light spots from the table to the breadbasket with hollowed white walls.",
"Move both the soft bread with golden edges and the spongy golden brown bread inside the white basket with beige bottom.",
"Place the soft bread with golden edges and ... |
8 | ["Lift the brown loaf and the golden bread and place them in the breadbasket made of woven wood.","P(...TRUNCATED) | ["Transfer the rounded oval bread and the bread loaf medium-sized from the table to the round beige (...TRUNCATED) |
10 | ["Use the right arm to place the golden bread in the breadbasket for baking.","Use the right arm to (...TRUNCATED) | ["Place the bread with golden crust into the beige breadbasket.","Pick up the bread with golden crus(...TRUNCATED) |
End of preview.
RoboTwin 2.0 — place_bread_basket expert demonstrations (500 episodes)
Self-collected expert demonstrations for the place_bread_basket task in
RoboTwin 2.0, recorded with the built-in
expert planner (play_once). Intended for behavior-cloning / VLA fine-tuning.
Collection settings
Collected with a custom task config derived from the official demo_clean /
demo_randomized presets. The domain randomization was narrowed to match a
deployment distribution that randomizes object pose and lighting only —
no distractor objects, no background texture randomization.
episode_num: 500
use_seed: false # seeds were searched for solvability, then replayed
save_freq: 15 # one frame per 15 control steps -> data at 15 Hz
embodiment: [aloha-agilex]
language_num: 100
domain_randomization:
random_background: false # keep the standard textured-light scene
cluttered_table: false # no distractor objects on the table
clean_background_rate: 1
random_head_camera_dis: 0 # head camera pose fixed
random_table_height: 0
random_light: true # per-episode random light color
crazy_random_light_rate: 0 # no per-frame light flicker
Per-episode randomization (driven by the episode seed) therefore covers:
| Axis | Randomized |
|---|---|
| Bread x/y position | yes |
| Bread yaw | yes, up to 45° |
| Bread mesh variant | yes, 5 variants |
| Basket mesh variant | yes, 5 variants |
| Light color | yes |
| Background texture | no |
| Table height | no |
| Head camera pose | no |
| Distractor objects | no |
Layout
data/episode_XXXXXXX.hdf5 # 500 files, ~17 MB each
video/episode_XXXXXXX.mp4 # 500 files, head-camera only, for inspection
instruction/episode_XXXXXXX.json # per-episode natural-language instructions
scene_info.json # per-episode object variants + language slots
seed.txt # 500 episode seeds, in completion order
Episode indices are contiguous 0..499; seed.txt maps episode index to seed.
HDF5 schema
state/ left_arm_joint_states (T, 6) right_arm_joint_states (T, 6)
left_ee_joint_states (T, 1) right_ee_joint_states (T, 1)
left_ee_poses (T, 7) right_ee_poses (T, 7)
action/ ... same keys, shifted by one timestep (t+1)
vision/ cam_head, cam_left_wrist, cam_right_wrist, cam_third_view
colors (T,) JPEG-encoded RGB bytes
shape (3,) decoded frame shape
intrinsic_matrix (T, 3, 3)
extrinsics_matrix (T, 4, 4)
additional_info/frequency scalar = 15 (Hz)
instructions scalar JSON list of instruction strings
Note vision/ carries four camera streams. The .mp4 under video/ is
only the head camera and exists for human inspection — training consumes the
HDF5, which is self-contained.
Statistics
episodes 500 (0 failures, 0 corrupt files)
frames total 114,929
frames per ep min 157 / mean 230 / max 335
arm usage (measured from joint motion):
dual-arm 237 (47%)
right arm only 144 (29%)
left arm only 119 (24%)
bread count per episode:
1 bread 146 (29%)
2 breads 354 (71%)
The task decides how many arms to use from the object layout: a single bread, or two breads on the same side, is handled by one arm; two breads on opposite sides are handled by both arms simultaneously. The bread count is a byproduct of placement feasibility, not a designed 50/50 split.
Known issue: {a} language slot
scene_info.json records an {a} slot naming the arm used. In single-bread
episodes it is computed after the bread has been moved, so it reads the
post-move position and is frequently wrong (empirically ~50/50 right vs left
in episodes labelled right).
This comes from upstream RoboTwin (envs/place_bread_basket.py, single-bread
branch: remove_bread(0, 0) executes before the arm_info = ... assignment).
A subset of the instruction strings that embed {a} — templates such as
"Use {a} to place {B} in {A}." — are affected. The trajectory data itself is
unaffected.
Usage
import h5py
with h5py.File("data/episode_0000000.hdf5", "r") as f:
obs_joints = f["state/left_arm_joint_states"][:]
actions = f["action/right_ee_poses"][:]
rgb = f["vision/cam_head/colors"][0] # JPEG bytes, decode to view
freq = f["additional_info/frequency"][()] # 15
Conversion to LeRobot format is available in the upstream repo:
XPolicyLab/scripts/transform_lerobot_v21_format.py (and v30).
Attribution
Collected with RoboTwin 2.0 (MIT). Assets (bread, basket meshes, background textures) come from the RoboTwin 2.0 asset release and remain under their original terms.
- Downloads last month
- 407