메뉴
Use when editing src/explainability/gradcam.py, adding a new model to MODEL_REGISTRY, or asked about the Grad-CAM panel in explainability reports for the corn leaf disease project.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Use when reading or editing src/data/*, src/config.py, src/models/input_sizes.py, or config/dataset.yaml - covers image loading, dataset/output root resolution, target_size convention, and minority-class derivation for the corn leaf disease project.
Use when running or editing scripts/pipeline/explain_lime.py, scripts/pipeline/explain_report.py, or scripts/checks/lime_stability.py, or when asked about LIME visual reports, fidelity, or error analysis for the corn leaf disease project.
Use when reading or editing src/training/*, scripts/pipeline/train.py, or scripts/pipeline/train_baselines.py - covers class-balancing via WeightedRandomSampler, shared training utilities, and run/output versioning for the corn leaf disease project.
SOC 직업 분류 기준
| name | corn-gradcam |
| description | Use when editing src/explainability/gradcam.py, adding a new model to MODEL_REGISTRY, or asked about the Grad-CAM panel in explainability reports for the corn leaf disease project. |
src/explainability/gradcam.py añade un 4to panel opcional (config/dataset.yaml -> gradcam.enabled) vía hooks nativos de PyTorch, sin dependencias nuevas.
GRADCAM_TARGET_LAYERS mapea cada modelo del registry a su última capa con salida espacial.
Al añadir un modelo nuevo a MODEL_REGISTRY, agregar también su entrada en GRADCAM_TARGET_LAYERS - si no, Grad-CAM se omite con warning (fallback a 3 paneles).
