gaponet-data

name	gaponet-data
description	Run GapONet data preparation. Use when the user invokes /gaponet-data or $gaponet-data, or asks to process, validate, stage, copy, symlink, or prepare GapONet motion .npz data for training or inference.

GapONet Data

Run the GapONet data stage, not just a review. Work from the GapONet repository root, identified by scripts/prepare_gaponet_data.py, scripts/run_gaponet_job.py, and configs/.

Workflow

1. Resolve the repository root and work from it.
2. If the user supplied an input .npz, stage it into the default operator motion path:

python scripts/prepare_gaponet_data.py --input-data <input.npz> --output-dir ./runs/data --mode copy --profile operator --force --update-configs

3. If the user did not supply an input path, validate the configured/default data:

python scripts/prepare_gaponet_data.py --config configs/data_default.json

4. If the default data is missing, ask for the source .npz path. Do not invent or synthesize motion data.
5. Report the generated runs/data/data_manifest.json status, the staged output_data, and any validation errors.

Defaults

• Staged data: source/sim2real/sim2real/tasks/humanoid_operator/motions/motion_amass/edited_27dof/test.npz
• Data config: configs/data_default.json
• Manifest: runs/data/data_manifest.json
• Downstream configs updated by --update-configs: configs/deploy_default.json and input_package_example.json

Validation Profile

Use the operator profile by default. It requires:

• real_dof_positions
• real_dof_velocities
• real_dof_positions_cmd
• real_dof_torques
• joint_sequence
• payloads

Use --profile deploy only when the user explicitly wants lightweight inference validation that does not require the operator training keys.

Completion

Consider the data stage complete only when data_manifest.json has "status": "success" and the staged .npz path exists. If Python lacks NumPy, tell the user to activate the GapONet environment before rerunning the data stage.