菜单
Use when navigating, reviewing, or changing the Sim2Real staged pipeline engine — stage map, preamble/inner/outer/finalize entrypoints, K8s sibling jobs, and S3 artifact contracts.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when ingesting, validating, curating, or querying production sensor data as a versioned dataset-of-record, or wiring the dataset-ingest-curate workflow.
Use when generating adversarial scenarios via RL, ranking mined failures of a policy-under-test, or wiring the adversarial-scenario-hardening workflow.
Use when navigating the Cosmos3 integration in NPA or locating upstream Cosmos3 framework files, defaults, scripts, configs, recipes, and docs.
Use when Cosmos3 setup, fetch, inference, CUDA, uv, Docker, Hugging Face, GitHub, NGC, or checkpoint staging fails in NPA or in an upstream Cosmos framework checkout.
Use when setting up Cosmos3 access through NPA, checking source or Hugging Face reachability, staging the public Cosmos3 framework and checkpoint cache, or deciding which NPA workflow to use before inference.
Use for Claude Code reviews that need robotics, simulation, GPU-routing, sim-to-real, or BDD100K pipeline domain context.
| name | sim2real-engine |
| description | Use when navigating, reviewing, or changing the Sim2Real staged pipeline engine — stage map, preamble/inner/outer/finalize entrypoints, K8s sibling jobs, and S3 artifact contracts. |
Load this skill when you need the canonical 14-stage map for the VLM-to-RL
Sim2Real loop, not the older generic sim-to-real workflow YAMLs. Use it for
engine edits, monitor/status alignment, walkthrough docs, and debugging sibling
K8s jobs (s2r-*).
| File | Role |
|---|---|
npa/src/npa/workflows/sim2real/engine.py | Stage glue, K8s siblings, inner/outer/finalize |
npa/src/npa/workflows/sim2real_stages.py | Mandatory preamble stages 3–7 helpers |
npa/src/npa/workflows/sim2real/monitor.py | _STAGE_SPECS S3 marker rules for live status |
npa/src/npa/workflows/sim2real/runner.py | Sim2RealWorkflow orchestration CLI entry |
npa/src/npa/workflows/sim2real_assets.py | Stage 2 assets consumption |
| Stage | Monitor name | Entrypoint | Primary artifacts |
|---|---|---|---|
| 1 | stage_01_trigger | run_preamble | stage_01_trigger/trigger.json |
| 2 | stage_02_assets | run_preamble → run_assets_stage | stage_02_assets/consumed_*_spec.json |
| 3 | stage_03_augment | run_preamble → run_augment_stage | augment/manifest.json |
| 4 | stage_04_envs_raw | run_envgen_split_stage | envs/raw/ |
| 5 | stage_05_envs_train | run_envgen_split_stage | envs/train/envs.jsonl |
| 6 | stage_06_tokens | run_envgen_split_stage | tokens/manifest.json |
| 7 | stage_07_actions_train | run_inner_loop → run_policy_rollouts | actions/train/ |
| 8 | stage_08_vlm_eval_train | run_inner_loop → evaluate_rollout_with_vlm | vlm_eval/train/ |
| 9 | stage_09_training_signal | run_inner_loop (signal + trainer) | training_signal/train/ |
| 10 | stage_10_eval_heldout | run_single_outer_iteration → run_heldout_eval | eval/heldout/report.json |
| 11 | stage_11_outer_loop | run_single_outer_iteration → threshold_decision | outer_loop/decision.json |
| 12 | stage_12_external_validation_stub | run_finalize | stage_12_external_validation/external_stub.json |
| 13 | stage_13_retrigger | run_finalize | stage_13_retrigger/retrigger.json |
| 14 | stage_14_rerun_viz | run_finalize → _run_sim2real_viz_stage | reports/sim2real.rrd |
run_preamble — trigger, assets, augment, envgen split, tokens.run_single_outer_iteration — one inner loop (7–9),
held-out eval (10), threshold decision (11). Repeats for outer_iterations.run_finalize — external-validation stub,
loop-of-loops retrigger record, local Rerun .rrd, sim2real-report.json,
optional S3 upload.Stages 4–6 share one component name in monitor: stage_04_06_env_gen_split_tokens.
infer_from_later=True); early
stages may show PENDING while later stages succeed — trust current_stage.run_id.npa/.venv/bin/ruff check npa/src/npa/workflows/sim2real/engine.py
npa/.venv/bin/python -m pytest npa/tests/workflows/sim2real/ -q --tb=no