The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "tsfile/tsfile_py_cpp.pyx", line 567, in tsfile.tsfile_py_cpp.tsfile_reader_new_c
tsfile.exceptions.FileOpenError: 28:
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/tsfile/tsfile.py", line 271, in _split_generators
scan = self._scan_metadata(all_files)
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/tsfile/tsfile.py", line 318, in _scan_metadata
with self._open_reader(file) as reader:
~~~~~~~~~~~~~~~~~^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/tsfile/tsfile.py", line 742, in _open_reader
return TsFileReader(file)
File "tsfile/tsfile_reader.pyx", line 323, in tsfile.tsfile_reader.TsFileReaderPy.__init__
SystemError: <class '_weakrefset.WeakSet'> returned a result with an exception set
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/split_names.py", line 71, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.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.
G1 WBT Inspire Put Drinks Into Fridge — TsFile edition
This repository is a time-series conversion of the LeRobot dataset unitreerobotics/G1_WBT_Inspire_Put_Drinks_Into_Fridge. The robot performs the task “put drinks into fridge and close the fridge” using a Unitree G1 with Inspire hands.
Source and attribution
- Dataset owner and author: Unitree Robotics (
unitreerobotics). - Source repository contributor: karthus198 (the Hugging Face file history lists this account as the contributor of the dataset files).
- Framework: LeRobot.
- License: Apache License 2.0.
- Paper/citation: the original dataset card does not provide a paper, BibTeX citation, or formal citation text. Cite the original Unitree Robotics dataset URL and this conversion when using the data.
Dataset summary
| Property | Value |
|---|---|
| Split | train |
| Episodes / devices | 300 |
| Rows / frames | 187,657 |
| Tasks | 1 |
| Sampling rate | 30 Hz |
| Source frame Parquet files | 2 |
| Converted TsFile files | 1 merged file |
| Source Parquet size | 106,247,696 bytes |
| Converted TsFile size | 71,103,021 bytes |
| TsFile / source Parquet ratio | 0.6692 |
The source split covers episodes 0:300. Both source Parquet shards are merged into
data/unitreerobotics_g1_wbt_inspire_put_drinks_into_fridge.tsfile.
TsFile schema
episode_index and task_index are table-model TAG columns. Together they identify a TsFile device; the 300 source episodes therefore remain queryable as 300 devices. frame_index and sample_index are ordinary FIELD columns.
| Source feature | Converted columns | Type | Role |
|---|---|---|---|
timestamp |
Time |
INT64 milliseconds | TIME |
episode_index |
episode_index |
STRING-backed TAG | TAG/device |
task_index |
task_index |
STRING-backed TAG | TAG/device |
frame_index |
frame_index |
INT64 | FIELD |
index |
sample_index |
INT64 | FIELD |
observation.state.ee_state (12) |
observation_state_ee_state_0 … _11 |
FLOAT | FIELD |
observation.state.hand_state (12) |
observation_state_hand_state_0 … _11 |
FLOAT | FIELD |
observation.state.robot_q_current (36) |
observation_state_robot_q_current_0 … _35 |
FLOAT | FIELD |
action.ee_action (12) |
action_ee_action_0 … _11 |
FLOAT | FIELD |
action.hand_cmd (12) |
action_hand_cmd_0 … _11 |
FLOAT | FIELD |
action.robot_q_desired (36) |
action_robot_q_desired_0 … _35 |
FLOAT | FIELD |
The end-effector fields contain the concatenated left/right end-effector poses. Hand-state and hand-command fields contain the 12 Inspire-hand values. Robot configuration fields contain root position, root quaternion, and 29 joint values.
Time, flattening, and storage policy
Time = round(timestamp * 1000)in milliseconds. Source timestamps restart at zero for each episode.- The source
timestampis omitted after conversion because its information is represented byTime; equivalently,timestamp ≈ Time / 1000subject to millisecond rounding. indexis renamed tosample_index;frame_index,episode_index, andtask_indexare preserved without changing their values.- Fixed-size vectors are flattened row-major into scalar FLOAT measurements. Periods in source names become underscores and element indices are appended.
- Rows are sorted by
episode_index,task_index, thenTime. Validation found no duplicate(episode_index, task_index, Time)key and Time is monotonic within every episode.
The compact codec profile is:
| TsFile type | Encoding | Compression |
|---|---|---|
| FLOAT / DOUBLE | GORILLA | LZ4 |
| INT32 / INT64 | TS_2DIFF | LZ4 |
| Time | TS_2DIFF | LZ4 |
| BOOLEAN | RLE | LZ4 (no BOOLEAN fields occur in this dataset) |
| TAG | TsFile table device/tag mechanism | TsFile tag storage |
Videos
Videos are not included in this TsFile repository. Use the original dataset at
the Hugging Face videos/ tree.
The four original video features are:
videos/observation.images.head_stereo_left/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4videos/observation.images.head_stereo_right/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4videos/observation.images.wrist_left/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4videos/observation.images.wrist_right/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
Each stream is 640×480, 30 fps, AV1, without audio. Use the unchanged episode_index and frame_index plus the original meta/episodes/ records to align TsFile rows with source video frames.
Reading the converted file
from tsfile import TsFileReader
path = "data/unitreerobotics_g1_wbt_inspire_put_drinks_into_fridge.tsfile"
reader = TsFileReader(path)
print(reader.get_all_table_schemas())
The table name is unitreerobotics_g1_wbt_inspire_put_drinks_into_fridge. Conversion scripts and validation reports are intentionally kept local and are not part of the upload set.
- Downloads last month
- 23