skills.rest
qinyan-ai
qinyan-aiCOMMUNITY
Explore all skills by this author →
GitHub

timesfm-forecasting

Forecast univariate time series with zero-shot TimesFM models from Hugging Face.

48|6|Updated Mar 9, 2026
Search Tags:#forecasting#time-series#preflight#timesfm#quantiles#univariate
One-click install
npx skills add https://github.com/qinyan-ai/qinyan-academic-skills --skill timesfm-forecasting-qinyan-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: timesfm-forecasting
Source: https://github.com/qinyan-ai/qinyan-academic-skills/tree/main/skills/09-%E6%9C%BA%E5%99%A8%E5%AD%A6%E4%B9%A0%E4%B8%8E%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD/timesfm-forecasting
Command: npx skills add https://github.com/qinyan-ai/qinyan-academic-skills --skill timesfm-forecasting-qinyan-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires torch, timesfm, pandas, numpy, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Enables zero-shot forecasting of univariate time series using Google's TimesFM foundation model, delivering both point forecasts and calibrated prediction intervals without training a custom model.

Core Features & Use Cases

  • Zero-shot forecasting: forecast any univariate series without model fine-tuning.
  • Input flexibility: accepts CSV, DataFrame, or 1-D arrays; supports batched forecasting for multiple series.
  • Prediction intervals: returns quantile forecasts (e.g., 10th–90th percentile) to form prediction intervals.
  • Preflight checks: includes a mandatory system checker to verify RAM/GPU and disk space before loading weights.
  • Modular workflow: guidance and examples cover data preparation, model loading, and evaluation workflows.

Quick Start

Provide your time-series data as a list of 1-D arrays and specify a forecast horizon to obtain point forecasts and prediction intervals.

Frequently Asked Questions about timesfm-forecasting

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I do zero-shot time-series forecasting without training a model?▼

Zero-shot time-series forecasting is possible using the TimesFM foundation model to predict univariate series directly. You can supply inputs as CSV, DataFrame, or 1-D arrays to generate point forecasts without any model fine-tuning.

Can I get prediction intervals for univariate time series forecasts?▼

Yes, prediction intervals are generated by returning quantile forecasts, such as the 10th to 90th percentiles. This allows you to form calibrated prediction intervals alongside your standard point forecasts.

What Python environment do I need to run TimesFM for forecasting?▼

Running TimesFM requires Python 3.10 or higher with torch, pandas, and numpy installed. A mandatory preflight system checker verifies available RAM, GPU, and disk space before loading the TimesFM checkpoint weights.

How do I forecast multiple univariate time series in a batch?▼

Batch forecasting is supported by providing your multiple univariate time series as a list of 1-D arrays. You must specify a forecast horizon and use a ForecastConfig to compile the model before generating the batch forecasts.

What is the best way to prepare raw 1-D array data for TimesFM forecasting?▼

To prepare raw 1-D array data for TimesFM forecasting, ensure your inputs are structured as a list of 1-D arrays or converted into a DataFrame. The modular workflow then handles model loading, compilation via ForecastConfig, and evaluation.

Related Skills

Explore More Data & Analytics Skills→
vasilyu1983vasilyu1983
73

ai-ml-timeseries

Forecast time series data with LightGBM, Transformers, and SHAP.

Community
Advanced
AlterLab-IEUAlterLab-IEU
58

alterlab-timesfm

Forecast univariate time-series data with zero-shot TimesFM point and quantile forecasts.

Community
Advanced
crazymsncrazymsn
22

timesfm-forecasting

Forecast univariate time-series with TimesFM from CSV, DataFrame, or arrays.

Community
Advanced
Ritabrata-ChakrabortyRitabrata-Chakraborty

timesfm-forecasting

Generate zero-shot probabilistic forecasts for univariate time series with TimesFM.

Community
Advanced