alphaloong_20251228 TsFile

This dataset contains a TsFile conversion of the Hugging Face dataset EvanLam/alphaloong_20251228.

Modalities: Time-series. The original dataset is a LeRobot v2.1 robot dataset for robot type W1. Source metadata reports 254 episodes, 191,446 frames, 1 task, 1 data chunk, 18 fps, and 762 videos.

Converted Files

• data/alphaloong_20251228.tsfile: numeric robot frame data converted to one TsFile table.
• meta/: source metadata mirrored from the original dataset, with meta/info.json updated to describe the converted TsFile artifact.

Videos are not included in this repository. They remain in the original Hugging Face dataset under videos/.

Schema

The TsFile table is alphaloong_20251228.

• Time: millisecond timestamp synthesized as round(timestamp * 1000).
• TAG columns: episode_index, task_index.
• FIELD columns: frame_index, sample_index, observation_state_0 through observation_state_16, and action_0 through action_16.

Source vector columns are flattened as scalar FLOAT fields:

• observation.state[17] -> observation_state_0..observation_state_16
• action[17] -> action_0..action_16

The source timestamp column is not retained as a separate field because it is represented by the TsFile Time column. The source index column is renamed to sample_index. No numeric time-series rows are intentionally dropped.

Validation

The converted TsFile was validated locally:

• Staged Parquet rows: 191,446
• TsFile metadata rows: 191,446
• TsFile readback rows: 191,446
• Output file: data/alphaloong_20251228.tsfile

Read Example

from tsfile import TsFileReader

path = "data/alphaloong_20251228.tsfile"
with TsFileReader(path) as reader:
    schemas = reader.get_all_table_schemas()
    print(schemas.keys())