Hugging Face's logo

Datasets:
jzpeng
/
yzw_so100_stack_cube_v4

Tasks:
Robotics
Modalities:
Time-series
Size:
10K<n<100K
Tags:
LeRobot
so100
stack_cube
v4
tsfile
timeseries
License:
Dataset card Files
xet
 Community
yzw_so100_stack_cube_v4 / README.md
jzpeng's picture
jzpeng
Add TsFile (converted from DorayakiLin/yzw_so100_stack_cube_v4)
d557d33 verified
preview code
|
Raw
History Blame Contribute Delete
4.13 kB
metadata 
license: apache-2.0
task_categories:
  - robotics
tags:
  - LeRobot
  - so100
  - stack_cube
  - v4
  - tsfile
  - timeseries
  - time-series
  - robotics
  - modality:timeseries
size_categories:
  - 10K<n<100K
modality:
  - timeseries
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/yzw_so100_stack_cube_v4_train.tsfile

YZW SO100 Stack Cube V4 (TsFile)

Converted from DorayakiLin/yzw_so100_stack_cube_v4 at pinned revision da7c2ecf66ab30bff1ee455a83efc83b37a7781c. Modalities: Time-series.

Dataset description

This LeRobot v2.1 SO100 dataset records demonstrations of picking up a red cube and stacking it on a green cube.

The pinned task instruction is: “Pick up the red cube and stack it on the green cube.”

  • Source repository owner/publisher: DorayakiLin
  • License: apache-2.0 (declared by the source card).
  • Paper/homepage/citation: no completed paper, homepage, or citation is documented in the pinned source card unless linked above.

Source metadata note: The source README embeds an older 20-episode/9,692-frame snapshot. Counts here use the pinned meta/info.json, 200 source episode Parquets, staged rows, and TsFile metadata.

Dataset Scale

Split Episodes Tasks Source trajectory Parquets TsFile rows Sampling rate TsFile files/shards
train 200 1 200 95,833 30 Hz 1

The staged Parquet has 95,833 rows and 17 columns including Time; TsFile chunk metadata independently reports the same 95,833 rows.

TsFile schema

Column Role TsFile type Observed/source range
Time TIME INT64 0–20,333 ms; restarts per episode
episode_index TAG STRING source integer 0–199
task_index TAG STRING source integer 0–0
frame_index FIELD INT64 0–610 within an episode
sample_index FIELD INT64 0–95,832 globally

Exact remaining FIELD names/ranges and imported types:

  • action_0–action_5 (FLOAT): main shoulder pan/lift, elbow flex, wrist flex/roll, and gripper action
  • observation_state_0–observation_state_5 (FLOAT): the corresponding SO100 state

Conversion

  • All source episodes in the train split are merged into data/yzw_so100_stack_cube_v4_train.tsfile; episode_index and task_index are TAG dimensions.
  • Time = round(timestamp * 1000) in milliseconds. The source timestamp column is omitted because it is redundant with Time / 1000 seconds.
  • frame_index is retained. Source index is retained as sample_index.
  • Every vector is fully flattened: the complete source name is kept, . becomes _, and element indices are appended. Float vectors are imported as single-precision FLOAT fields.
  • Other scalar source columns shown above are retained; no trajectory rows are intentionally dropped.
  • Source metadata is mirrored for publication, with copied meta/info.json rewritten to describe the converted data path and conversion semantics.

Video policy

The pinned metadata declares 600 source videos across wrist, top, and third camera streams. MP4 files were not downloaded or uploaded; they remain in the pinned source videos/ tree.

Minimal read example 

from tsfile import TsFileReader

reader = TsFileReader("data/yzw_so100_stack_cube_v4_train.tsfile")
print(reader.get_all_table_schemas().keys())
reader.close()

Source and provenance