universal-tasks
Use and debug TimeMixer imputation, anomaly detection, and classification workflows.
- License 许可
- Apache-2.0
- 最近更新
- 2026-10-08
- Stars
- ★ 11
- Forks
- 3
01 前言 Front Matter
- name
- universal-tasks
- license
- Apache 2.0
- metadata
- { "disco-role": "operating" }
- description
- Use and debug TimeMixer imputation, anomaly detection, and classification workflows.
- disable-model-invocation
- true
02 资源文件 2
- 参考文档references/task-workflows.md6.8 KB
- 参考文档references/troubleshooting.md4.2 KB
03 SKILL.md 正文
TimeMixer Universal Tasks
Use this sub-skill for TimeMixer's non-forecast branches: imputation, anomaly detection, and classification.
Route here for
- Building or checking
run.pycommands for imputation, anomaly detection, or classification, using the explicit helper path/path/to/timemixer-skill/sub-skills/universal-tasks/scripts/build_universal_task_command.py. - Understanding task-specific args such as
mask_rate,anomaly_ratio, UEA padding, andchannel_independence. - Locating result files, saved metrics, and checkpoint paths for the non-forecast branches.
- Debugging dataset-name, mask, threshold, or classification shape issues that are specific to these tasks.
The helper prints commands only; run generated
run.pycommands from/path/to/TimeMixer-checkout, not from the skill directory.
Safe defaults
- Use the command builder to assemble a command before any training run. Its source-compatible downsampling defaults are
--down-sampling-layers 1 --down-sampling-window 2; overrides must keep layers >= 1, window >= 2, and the smallest scale non-empty. - M4 imputation is rejected by the builder because this source M4 loader cannot safely use the universal branch's
pred_len=0; use the forecasting M4 presets instead. - Keep
c_outaligned with the input channel count for imputation and anomaly detection. - For multi-feature UEA classification, prefer
channel_independence=0. - Treat
anomaly_ratioas a percent value in the source threshold calculation. - Do not launch benchmark-scale training or dataset downloads from this sub-skill.
Route elsewhere
- Forecasting benchmark scripts, custom forecast presets, or M4/PEMS/ECL/ETT/Solar recipes:
forecasting-experiments. - Dataset file layout, CSV/NPY/TS placement, or validation of raw inputs:
data-preparation. - TimeMixer tensor internals, PDM/FMM shapes, or decomposition/downsampling behavior:
model-architecture.
Task map
| Task | Source branch | Typical data key | Main metric | Main outputs |
|---|---|---|---|---|
| Imputation | Exp_Imputation |
ETTh1, ETTh2, ETTm1, ETTm2, custom, m4, PEMS, Solar |
Masked-value MSELoss during training; MAE/MSE/RMSE/MAPE/MSPE on test |
result_imputation.txt, results/<setting>/metrics.npy, pred.npy, true.npy |
| Anomaly detection | Exp_Anomaly_Detection |
PSM, MSL, SMAP, SMD, SWAT |
Accuracy, precision, recall, F-score after thresholding and adjustment | result_anomaly_detection.txt |
| Classification | Exp_Classification |
UEA |
Cross-entropy loss; accuracy for validation/test | results/<setting>/result_classification.txt |
Bundled assets
references/task-workflows.mdreferences/troubleshooting.mdscripts/build_universal_task_command.py
Safe defaults
- Use the command builder to assemble a command before any training run.
- Keep
c_outaligned with the input channel count for imputation and anomaly detection. - For multi-feature UEA classification, prefer
channel_independence=0. - Treat
anomaly_ratioas a percent value in the source threshold calculation. - Do not launch benchmark-scale training or dataset downloads from this sub-skill.
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.