training-and-evaluation
Use for Anomalib's Engine fit/train/test/validate flow, metrics, callbacks, preprocessing, postprocessing, loggers, visualization, and training-mode guidance while keeping CLI and deployment internals out of scope.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- training-and-evaluation
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Use for Anomalib's Engine fit/train/test/validate flow, metrics, callbacks, preprocessing, postprocessing, loggers, visualization, and training-mode guidance while keeping CLI and deployment internals out of scope.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/callbacks-and-logging.md4.3 KB
- 参考文档references/engine-and-metrics.md7.4 KB
- 参考文档references/troubleshooting.md4.5 KB
03 SKILL.md 正文
Training and Evaluation
Use this sub-skill when the user needs to configure, explain, or debug Anomalib's training/evaluation loop, metric collection, callback order, preprocessing/postprocessing, experiment logging, or visualization.
Read first
references/engine-and-metrics.mdreferences/callbacks-and-logging.mdreferences/troubleshooting.mdscripts/training-modes.pyscripts/evaluation-smoke.py
Safe operating boundary
Owns the core training/evaluation surface under:
src/anomalib/enginesrc/anomalib/callbackssrc/anomalib/metricssrc/anomalib/pre_processingsrc/anomalib/post_processingsrc/anomalib/loggerssrc/anomalib/visualization
Covers:
Engine.fit,Engine.train,Engine.validate,Engine.test,Engine.predict,Engine.from_configAnomalibModule.configure_callbacksget_callbacksEvaluator,AnomalibMetric,PostProcessor,PreProcessor,ImageVisualizer- training-mode guidance such as
barebones, checkpointing, and validation-first zero/few-shot behavior
Excludes:
- CLI parsing details
- export/deployment internals
- pipeline orchestration
Treat optional logger packages and GPU-only device checks as non-blocking.
Quick routing
| User need | Do this |
|---|---|
| "How do I wire training, validation, and test?" | Read references/engine-and-metrics.md. |
| "Why do my callbacks or loggers behave strangely?" | Read references/callbacks-and-logging.md. |
| "Why are metrics missing, duplicated, or empty?" | Read references/troubleshooting.md. |
| "Can I run a CPU-only smoke for barebones mode?" | Run scripts/training-modes.py. |
| "Can I sanity-check metrics, preprocessing, postprocessing, and visualization?" | Run scripts/evaluation-smoke.py. |
Operating notes
- Model components are usually chained as
pre_processor -> model -> post_processor, withevaluatorandvisualizerattached as callbacks. - Validation populates normalization and threshold state for
PostProcessorand related metrics. - The default evaluator uses prefixed image/pixel metrics so image-level and pixel-level results do not collide.
- Barebones mode removes Lightning logging overhead; use Anomalib's evaluator if you still need returned metrics.
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.