技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
chronos-bolt-and-original【数据与研究】Use Chronos-Bolt and original Chronos/T5 pipelines for univariate forecasting, quantiles, samples, DataFrames, embeddings, fev, and model-family selection.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when the task is about the Chronos-Bolt or original Chronos/T5 families from the chronos-forecasting package:
predict or predict_quantiles;predict_df;embed;predict_fev bridge for fev tasks;future_df, cross_learning, Chronos-2 fine-tuning, or Chronos-2 deployment choices, route to ../chronos-2-forecasting/.../data-formats-and-validation/.../training-evaluation-deployment/.references/api-reference.md — exact public signatures, tensor shapes, forecast-type differences, loading behavior, and family properties.references/workflows.md — practical loading, tensor/list prediction, quantile/mean, DataFrame, embeddings, fev, and model-selection recipes.references/troubleshooting.md — shape, padding/NaN, sample-vs-quantile, prediction-length, S3/HF cache, dtype/device, and dependency fixes.scripts/bolt_original_smoke.py defaults to import/signature inspection when no model is provided.--model-id-or-path is supplied; non-local Hugging Face or S3 identifiers additionally require --allow-remote.num_samples, temperature, top_k, top_p), or compatibility with original Chronos behavior.