THULab/macrodata_aloha_static_battery_ep005_009

ALOHA Static Battery Episodes 005-009 TsFile

Apache TsFile representation of `macrodata/aloha_static_battery_ep005_009`, a LeRobot v3 ALOHA manipulation dataset.

Source

• —Repository owner: Macrodata Labs
• —Repository contributor and uploader: hynky
• —License: Apache-2.0
• —Task: Place the battery into the slot of the remote controller.
• —Split: train; 5 episodes; 3,000 frame rows; 1 task; 50 fps
• —Numeric source layout: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• —The source card provides no paper, homepage, separate author list, or completed citation.

Data layout

• —TsFile: data/macrodata_aloha_static_battery_ep005_009.tsfile (130,413 bytes)
• —Table: macrodata_aloha_static_battery_ep005_009
• —Rows: 3,000; episode TAG values: 0 through 4
• —Time precision: milliseconds

Time = round(timestamp * 1000) and restarts at zero for each episode. The redundant source timestamp is omitted because it equals Time / 1000 seconds. Source index is named sample_index, and frame_index is retained.

The 14-element action and observation.state vectors are flattened to scalar FLOAT fields. No rows or numeric action/state dimensions are omitted.

Videos

The four original AV1 streams are available in the source `videos/` tree:

• —videos/observation.images.cam_high/chunk-000/file-000.mp4
• —videos/observation.images.cam_left_wrist/chunk-000/file-000.mp4
• —videos/observation.images.cam_low/chunk-000/file-000.mp4
• —videos/observation.images.cam_right_wrist/chunk-000/file-000.mp4

Videos are not included here. Rows stay aligned with the original frames through episode_index and frame_index.

Read example

from tsfile import TsFileReader

reader = TsFileReader("data/macrodata_aloha_static_battery_ep005_009.tsfile")
with reader.query_table(
    "macrodata_aloha_static_battery_ep005_009",
    ["episode_index", "task_index", "frame_index", "next_done", "action_0"],
    batch_size=4096,
) as result:
    print(result.read_arrow_batch().to_pandas().head())
reader.close()

Please cite the original Hugging Face dataset and its repository owners/contributors when reusing the data.