model-specs
Use when asked about a catalog model's own static characteristics — architecture, dataset/training compatibility, serving runtime requirements — e.g. "what was DreamZero trained on" or "does pi0.5 need a fixed camera count." Not for live status, scaling, or calling a deployed model (see the models skill for that).
소스 정보
- 저장소
- redhat-et/physical-ai-skills
- 최근 소스 활동
- 2026년 8월 4일 22:21
- 감지된 SKILL.md 언어
- 영어
- 스타
- 0
- 포크
- 0
설치 방법
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
소스 파일 검토
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
파일 탐색기
4 개 파일SKILL.md 표시 중
SKILL.md
소스 지침 · 읽기 전용 미리보기- name
- model-specs
- description
- Use when asked about a catalog model's own static characteristics — architecture, dataset/training compatibility, serving runtime requirements — e.g. "what was DreamZero trained on" or "does pi0.5 need a fixed camera count." Not for live status, scaling, or calling a deployed model (see the models skill for that).
MODEL SPECS — this skill is knowledge-only: no cluster calls, no
scaling/inference actions. It's the reference for what a catalog model
*is*, as opposed to the models skill (what a deployed instance is
currently doing).
## Scripts
Run via the shell tool: `python3 "$SKILLS_ROOT/model-specs/scripts/get_model_reference.py" --model-name <name>`.
Each catalog model with a documented dataset/training profile has a file
under `references/<model_name>.md` (e.g. `references/pi05.md`,
`references/dreamzero.md`), keyed by the same catalog directory name used
elsewhere (platform/base/models/<model_name>/). A model with no file here
either has no fine-tuning recipe on this platform or hasn't been
documented yet — don't guess its specifics from general knowledge of the
base model.
pi0.5's reference is a Dimension/Priority checklist table matching the
datasets skill's own DATASET COMPATIBILITY CHECKLIST row for row — read
them side by side by row number when validating a candidate dataset.
DreamZero's is plain training-data provenance prose instead of a table,
since it's inference-only on this platform (no recipe to check a dataset
against).
GitHub에서 보기