Datasets:

Ada-4321
/

SkillWeaver-NIS

ArXiv:

Tags:

robotics

reinforcement-learning

manipulation

isaac-lab

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    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/webdataset/webdataset.py", line 80, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              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 68, 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.

SkillWeaver-NIS data

Data bundle for SkillWeaver-NIS, the Isaac Lab training environments of the Neural Interaction Skills (NIS) used in SkillWeaver (CoRL 2026, arXiv:2609.36171). It contains example assets, sampled initial arm configurations and the rl_games checkpoints of all released sub-policies.

Download

From the root of a SkillWeaver-NIS clone:

pip install -U huggingface_hub
hf download Ada-4321/SkillWeaver-NIS --repo-type dataset --local-dir data --exclude README.md .gitattributes
tar -xzf data/assets.tar.gz -C data && rm data/assets.tar.gz

The default paths in isaaclabenvs/config/config.template.toml point at this layout, so no configuration is needed afterwards.

The objects are packed in assets.tar.gz (0.8 GB), which extracts to assets/:

assets/
├── rigid/          # 153 objects in 61 categories: <scenario>/<category>/asset_XXXX/
└── arti/
    ├── prismatic/  # LIBERO cabinets 90011 and 90021, drawers 1-3, 5 sizes each
    └── revolute/   # LIBERO microwave 90031, door, 5 sizes
initial_pose/       # initial arm joint configurations, sampled per object position and table height
checkpoints/        # one rl_games checkpoint per sub-policy

Rigid objects: asset_XXXX_coacd/coacd_final.usd (visual mesh and convex-decomposition collision), asset_XXXX_pose.npz (stable resting poses on a table), asset_XXXX_pcd.npz and per-face point clouds, asset_XXXX.json (category, axes and size; handle side for objects with a handle) and grasp_pose.franka_panda.npz (Franka grasps, used by Place).
Articulated objects: USD and URDF per size variant, handle points, moving-link points, handle grasps (Open) and push poses (Close).

The rigid objects are the example assets of each rigid-object sub-policy, 20 per sub-policy, drawn from the set its checkpoint was trained on; the per-sub-policy lists are in data/asset_lists/ of the GitHub repository. The checkpoints were trained on larger object sets.

Checkpoint	Task ID
`pick_general.pth`	`SkillWeaver-Franka-Pick-General-v0`
`pick_upright.pth`	`SkillWeaver-Franka-Pick-Upright-v0`
`pick_concave.pth`	`SkillWeaver-Franka-Pick-Concave-v0`
`pick_concave_lying.pth`	`SkillWeaver-Franka-Pick-ConcaveLying-v0`
`pick_handle.pth`	`SkillWeaver-Franka-Pick-Handle-v0`
`place_general.pth`	`SkillWeaver-Franka-Place-General-v0`
`place_upright.pth`	`SkillWeaver-Franka-Place-Upright-v0`
`open_drawer.pth`	`SkillWeaver-Franka-Open-Drawer-v0`
`open_door.pth`	`SkillWeaver-Franka-Open-Door-v0`
`close_drawer.pth`	`SkillWeaver-Franka-Close-Drawer-v0`
`close_door.pth`	`SkillWeaver-Franka-Close-Door-v0`
`widowx_pick.pth`	`SkillWeaver-WidowX-Pick-v0`

The articulated objects are derived from LIBERO (MIT License).

Citation

@misc{zhu2026skillweaveragenticexplorationneural,
      title={SkillWeaver: Agentic Exploration over Neural Interaction Skills for Scalable Robot Data Generation},
      author={He Zhu and Lusen Zhao and Kwan Man Cheng and Su Li and Katerina Fragkiadaki},
      year={2026},
      eprint={2609.36171},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2609.36171},
}