training-configuration
Configure and launch nnU-Net v2 training runs after planning/preprocessing is complete: folds, configurations, resume/validation flags, devices/DDP, pretrained weights, trainer selection, checkpoints, logs, manual splits, and command matrices.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- training-configuration
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Configure and launch nnU-Net v2 training runs after planning/preprocessing is complete: folds, configurations, resume/validation flags, devices/DDP, pretrained weights, trainer selection, checkpoints, logs, manual splits, and command matrices.
- disable-model-invocation
- true
02 资源文件 3
- 参考文档references/cli-reference.md6.8 KB
- 参考文档references/training-workflows.md6.9 KB
- 参考文档references/troubleshooting.md7.5 KB
03 SKILL.md 正文
nnU-Net v2 Training Configuration
Use this sub-skill when the user needs to train nnU-Net v2 models on already planned and preprocessed data, choose folds/configurations, recover validation exports, resume runs, select devices, or understand training outputs.
Do not use this sub-skill for dataset conversion, planning, preprocessing, inference, best-configuration selection, ensembling, or implementing new trainer subclasses. For custom trainer code, route to customization-extension; this sub-skill only shows how to select trainer class names in training commands.
Fast Path
- Confirm
nnUNet_raw,nnUNet_preprocessed, andnnUNet_resultsare set and that the dataset has been planned/preprocessed. - Choose a trained configuration from the dataset plans, commonly
2d,3d_fullres,3d_lowres, or3d_cascade_fullres. - Train folds with
--npzif model selection or ensembling may be needed later:nnUNetv2_train DATASET001_Example 3d_fullres 0 --npz - Train all folds by repeating folds
0through4, or train one final model on all cases with foldallwhen cross-validation is not needed:nnUNetv2_train DATASET001_Example 3d_fullres all - For cascade training, complete
3d_lowresfirst, then train3d_cascade_fullres. - Inspect outputs under
nnUNet_results/DatasetXXX_Name/TRAINER__PLANS__CONFIGURATION/fold_X/.
Bundled References
references/cli-reference.md: exactnnUNetv2_trainarguments, Python API mapping, flags, device/DDP semantics, and output paths.references/training-workflows.md: common training matrices, cascade ordering, manual splits, pretraining/fine-tuning, logging, and command generation.references/troubleshooting.md: failure-mode triage for missing preprocessing, cascade dependencies,--npzrecovery, resume/validation choices, DDP, checkpoints, CPU/MPS expectations, and manual split placement.scripts/plan_training_matrix.py: prints command matrices for folds/configurations without running training.
Command Matrix Helper
Use the bundled helper to produce reviewable shell commands before launching expensive jobs:
python sub-skills/training-configuration/scripts/plan_training_matrix.py DATASET001_Example --configs 2d 3d_fullres --folds 0 1 2 3 4 --npz
It only prints commands. It does not inspect data, allocate GPUs, submit cluster jobs, or run nnUNetv2_train.
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.