timesfm-forecasting

Generates zero-shot time-series forecasts and prediction intervals from CSV, DataFrame, or array inputs using Google's TimesFM foundation model and torch library.

3|Updated Apr 17, 2026

Search Tags:#zero-shot #timesfm #univariate #timeseries forecasting #probability intervals

Ready to run

npx skills add https://github.com/RamanEbrahimi/raman-marketplace --skill timesfm-forecasting-ramanebrahimi

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: timesfm-forecasting
Source: https://github.com/RamanEbrahimi/raman-marketplace/tree/main/plugins/agentic-research/skills/scientific-skills/timesfm-forecasting
Command: npx skills add https://github.com/RamanEbrahimi/raman-marketplace --skill timesfm-forecasting-ramanebrahimi

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

This Skill provides zero-shot time series forecasting with Google's TimesFM foundation model, enabling accurate predictions without custom training. It's ideal for any univariate time series and offers probabilistic forecasts with prediction intervals.

Core Features & Use Cases

Zero-shot Forecasting: Utilizes TimesFM's pre-trained foundation model for accurate predictions on any univariate time series.
Probabilistic Forecasts: Provides calibrated quantile prediction intervals for robust confidence in forecasts.
Flexible Inputs: Accepts CSV, DataFrame, and array inputs for point forecasts and prediction intervals.
Preflight System Checker: Ensures adequate system resources before loading the model to prevent crashes.
Use Case: Forecast sales, energy consumption, or weather data without training a custom model.

Quick Start

Use the timesfm-forecasting skill to forecast the next 12 months of sales data in the 'sales_data.csv' file.

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 custom model?▼

Zero-shot time series forecasting is achievable using the TimesFM foundation model to predict univariate data without custom training. This skill leverages the pre-trained model to generate accurate forecasts directly on your data.

Can I generate prediction intervals and calibrated quantiles for univariate time series data?▼

Yes, probabilistic forecasting with calibrated quantile prediction intervals is fully supported. The skill generates point forecasts and robust confidence intervals for univariate time series to provide reliable forecast ranges.

Does TimesFM forecasting accept CSV files and DataFrames as input?▼

TimesFM forecasting accepts CSV files, Pandas DataFrames, and array inputs for flexible data ingestion. You can directly load your univariate time series data from these formats to generate predictions without extensive preprocessing.

Do I need PyTorch to run TimesFM for univariate time series predictions?▼

Yes, PyTorch is a required dependency to run TimesFM for univariate time series predictions. The skill requires both the TimesFM foundation model library and torch to perform zero-shot forecasting and generate output intervals.

What is the best way to forecast sales or energy consumption without historical model training?▼

The best way to forecast sales or energy consumption without training is using a pre-trained foundation model for zero-shot inference. TimesFM applies large-scale pre-training to accurately predict univariate time series across these domains.

How does the preflight system checker prevent crashes when loading the TimesFM model?▼

The preflight system checker verifies adequate system resources are available before loading the TimesFM model. This precaution ensures your environment can handle the large-scale foundation model, preventing crashes during the forecasting process.