Hugging Face's logo

Datasets:
jzpeng
/
humanego_serve_bread_lingbot_lerobot_with_latents

Modalities:
Time-series
Tags:
lerobot
robotics
robot-learning
imitation-learning
time-series
timeseries
Dataset card Files
xet
 Community
Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'embeddings' of the dataset.
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.

HumanEgo Serve Bread LingBot TsFile With Latents

This is an Apache TsFile conversion of Coffeecoderss/humanego_serve_bread_lingbot_lerobot_with_latents at pinned revision 18911305890ce1ca623e4399834b63e792ff7534. The source contains LeRobot demonstrations for the task pick up the bread and place it on the plate.

  • Modalities: Time-series
  • Source layout: two standalone LeRobot v2.1 roots (train and validation)
  • Trajectory rows: 41,603 train and 5,533 validation
  • Latent rows: 3,326,400 train and 430,080 validation
  • Tensor windows: 559 unique source tensor windows
  • Embeddings: 2 unique embeddings after exact logical-tensor SHA256 deduplication

Schema

Trajectory tables use Time = round(timestamp * 1000) milliseconds, with episode_index and task_index as TAGs. frame_index is retained, source index becomes sample_index, dotted names use underscores, and vector state/action values are flattened into FLOAT fields.

Latent tables use TAGs episode_index, window_start_frame, window_end_frame, spatial_y, and spatial_x. Their FLOAT fields are latent_0..latent_47 for the production source. Latent Time uses round(frame_ids[4 * latent_t] * 1000 / ori_fps) as a deterministic coordinate mapping, not a claim about encoder receptive-field centers.

The embedding table uses 64-value channel blocks. TAGs identify embedding ID, kind, and block; Time is the token ordinal; FIELDs are value_0..value_63.

Preservation and exclusions

BF16 tensor values were expanded exactly into FP32 FLOAT values. Repeated text and empty embeddings were stored once by logical-tensor SHA256, while manifest/conversion_manifest.json retains every source-file association and reconstruction mapping.

Videos and original PTH/PT archives are not uploaded. Videos remain in the pinned source repository; copied metadata retains the original video description.

Source & license

Minimal read example 

from tsfile import TsFileReader

reader = TsFileReader("data/trajectory_train.tsfile")
# Query with the Apache TsFile Python or Java SDK.
Downloads last month
58
Total file size:
1.18 GB