asap
Use ASAP to train humanoid policies, retarget SMPL motions, and run sim2sim or sim2real deployment workflows for humanoidverse-based robots.
- License 许可
- MIT
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- asap
- license
- MIT
- metadata
- { "disco-role": "operating" }
- description
- Use ASAP to train humanoid policies, retarget SMPL motions, and run sim2sim or sim2real deployment workflows for humanoidverse-based robots.
- disable-model-invocation
- true
02 资源文件 5
- 参考文档references/configuration-map.md2.4 KB
- 参考文档references/install-and-backends.md2.8 KB
- 参考文档references/repo-provenance.md1.7 KB
- 参考文档references/repo-routing-metadata.json508 B
- 参考文档references/troubleshooting.md2.4 KB
03 SKILL.md 正文
ASAP
Use this skill for the ASAP humanoid learning stack: training and evaluating policies in HumanoidVerse, retargeting SMPL/AMASS motion into robot motion files, and running sim2sim or sim2real deployment helpers.
Install and Import Check
Start with the bundled doctor script from the generated ASAP skill root, then add the backend-specific pieces you need.
python scripts/asap_doctor.py --help
python scripts/asap_doctor.py --repo-root <asap-checkout> --section core
For the common Python stack, install the editable packages plus the repo-local utilities you need:
pip install -e .
pip install -e isaac_utils
# optional for deployment helpers
pip install -e sim2real
Use a CUDA-capable PyTorch build for the training routes and the policy runtime routes that expect GPU acceleration. Install simulator- or hardware-specific SDKs separately when the doctor script reports them missing.
Read references/install-and-backends.md before choosing a simulator backend or attempting motion retargeting or robot deployment.
Route by Task
- Train, evaluate, or export humanoid policies: use
sub-skills/training-and-evaluation/SKILL.mdfor motion-tracking, locomotion, delta-action, checkpoint, and Hydra-config workflows. - Retarget SMPL or AMASS motion: use
sub-skills/motion-retargeting/SKILL.mdfor shape fitting, motion fitting, asset validation, and MuJoCo visualization. - Run sim2sim or sim2real deployment: use
sub-skills/sim2real-deployment/SKILL.mdfor MuJoCo playback, ROS2/Unitree bridges, policy toggles, joystick control, and data logging.
Common Starting Points
- If you just need a fast dependency or backend check, run
scripts/asap_doctor.pyfirst. - If a task fails with a Hydra or import-path error, read
references/troubleshooting.mdbefore editing commands. - If a task needs the current repository baseline, read
references/repo-provenance.mdto compare the checkout against the generated skill snapshot. - If you are unsure which config group to override, read
references/configuration-map.mdand then route to the relevant sub-skill.
Shared References
references/configuration-map.md— Hydra groups, common overrides, and output-directory conventions.references/install-and-backends.md— base package install, simulator backends, and external SDK prerequisites.references/troubleshooting.md— import, backend, config, and runtime failure patterns.references/repo-provenance.md— source commit and evidence baseline for refresh decisions.references/repo-routing-metadata.json— structured router placement metadata for import into repo-skills-router.
Boundaries
This is a runtime skill for the ASAP package, not a maintainer release guide. It does not cover unrelated repo maintenance, packaging policy, or non-ASAP robot stacks.
If a task is only about generic PyTorch, general RL theory, or a different simulator stack, route it elsewhere instead of forcing it through ASAP-specific guidance.
04 同仓库更多 Skills
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
obsidian-clipper-template-creator
Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.
python-api
Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP/mAP semantics, and source-style import diagnostics.
file-evaluation
Folder-based PASCAL VOC AP/mAP evaluation from ground-truth and detection text files.
object-detection-metrics
Guides legacy Object-Detection-Metrics PASCAL VOC AP/mAP evaluation from detection text folders and direct Python API classes.
nx-workspace-patterns
Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.