agentsclimarketplace

Data pipeline

Skill ZhangHanDong/dora-skills/skills/data-pipeline

"Use for data recording and replay questions with dora.From its SKILL.md

Install
npx -y skills add ZhangHanDong/dora-skills --skill data-pipeline

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 7 stars7 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.9 KB, as published. Nobody here has run it

Domain: Data Pipeline

Recording and replaying robot data with dora-rs

Overview

Dora supports data pipelines for:

  • Recording sensor data for training
  • Replaying recorded sessions
  • LeRobot integration for imitation learning
  • Dataset management

LeRobot Integration

Note: dora-lerobot is from a separate repository, not PyPI. Install from source:

git clone https://github.com/dora-rs/dora-lerobot
cd dora-lerobot
pip install -e dora_lerobot

Recording Data

nodes:
  # Camera
  - id: camera
    build: pip install opencv-video-capture
    path: opencv-video-capture
    inputs:
      tick: dora/timer/millis/33
    outputs:
      - image

  # Robot arm
  - id: arm
    build: pip install dora-rustypot
    path: dora-rustypot
    inputs:
      command: teleop/command
    outputs:
      - state
      - feedback

  # LeRobot recorder (install from dora-lerobot repo first)
  - id: recorder
    path: dora-lerobot-recorder
    inputs:
      image: camera/image
      state: arm/state
      action: teleop/command
    env:
      DATASET_NAME: my_robot_dataset
      EPISODE_INDEX: "0"

Replaying Data

nodes:
  # LeRobot replay (install from dora-lerobot repo first)
  - id: replay
    path: dora-lerobot-replay
    inputs:
      tick: dora/timer/millis/33
    outputs:
      - image
      - state
      - action
    env:
      DATASET_NAME: my_robot_dataset
      EPISODE_INDEX: "0"

  # Visualization
  - id: plot
    build: pip install dora-rerun
    path: dora-rerun
    inputs:
      image: replay/image

Data Collection Workflow

# 1. Start recording session
dora run record_dataflow.yml

# 2. Perform teleoperation
# 3. Press Ctrl+C to stop and save

# 4. Repeat for multiple episodes
EPISODE_INDEX=1 dora run record_dataflow.yml
EPISODE_INDEX=2 dora run record_dataflow.yml

# 5. Train policy
python train_policy.py --dataset my_robot_dataset

Available Hub Nodes

NodeInstallPurpose
dora-lerobot-recorderFrom dora-lerobot repoRecord data
dora-lerobot-replayFrom dora-lerobot repoReplay data
llama-factory-recorderpip install llama-factory-recorderRecord for LLM/VLM training
lerobot-dashboardpip install lerobot-dashboardPygame recording interface
dora-rdt-1bpip install dora-rdt-1bVLA policy inference

Training Pipeline

1. Collect Demonstrations

# Run teleoperation dataflow
dora run record_dataflow.yml

# Mark episodes
# Press 'n' for new episode, 'f' for failed

2. Convert to LeRobot Format

# The recorder automatically saves in LeRobot format
# Dataset saved to: ~/.lerobot/datasets/<dataset_name>

3. Train Policy

# Using LeRobot CLI
python lerobot/train.py \
    --dataset my_robot_dataset \
    --policy diffusion \
    --output_dir outputs/my_policy

4. Deploy Policy

nodes:
  - id: camera
    path: opencv-video-capture
    inputs:
      tick: dora/timer/millis/33
    outputs:
      - image

  - id: policy
    build: pip install dora-rdt-1b
    path: dora-rdt-1b
    inputs:
      image: camera/image
    outputs:
      - action

  - id: robot
    path: dora-piper
    inputs:
      joint_positions: policy/action

Related Skills

  • hub-recording - Detailed recording node documentation
  • hub-robot - Robot control nodes
  • domain-robot - Robot control patterns
  • domain-vision - Visual data processing
  • hub-nodes - All pre-built nodes

Keep looking

Skills are one crate of 610,677. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.