What are the prerequisites for using this skill?

You need Python 3.8+, the nixtla and pandas packages installed, and a valid NIXTLA_API_KEY configured in your environment variables.

How much data is required for TimeGPT fine-tuning?

For effective fine-tuning, Nixtla typically requires at least 100 observations per time series to ensure the model can capture domain-specific patterns.

How do I trigger the fine-tuning workflow?

You can activate the skill by telling Claude to 'fine-tune TimeGPT' or 'train a custom model' within your project directory.

Can I compare my fine-tuned model against the default TimeGPT?

Yes, the skill includes a comparison step that evaluates metrics like SMAPE to show the exact improvement of your fine-tuned model over the zero-shot baseline.

Where does the skill store training artifacts?

Configuration files are saved in the 'forecasting/' directory, while model IDs and performance metrics are stored in the 'artifacts/' and 'results/' subdirectories respectively.

Nixtla TimeGPT Fine-Tuning Lab

Name: Nixtla TimeGPT Fine-Tuning Lab
Author: intent-solutions-io

byintent-solutions-io

Ciencia de Datos y ML

Streamlines the fine-tuning of Nixtla TimeGPT models on custom datasets for high-precision, domain-specific forecasting.

Acerca de

The Nixtla TimeGPT Fine-Tuning Lab is a specialized skill designed to manage the end-to-end lifecycle of custom model training within Claude Code. It automates the complex steps of preparing time-series data for the Nixtla SDK, submitting fine-tuning jobs to the TimeGPT API, and monitoring progress. By providing integrated benchmarking tools, it allows developers to quantify the performance gains of specialized models over standard zero-shot forecasts, ensuring production-ready accuracy for industry-specific use cases like retail demand or energy consumption.

Características Principales

Automated dataset validation and Nixtla schema formatting
Real-time status monitoring for remote training jobs
Integrated error handling for API authentication and data volume requirements
Streamlined fine-tuning job submission and configuration management
Side-by-side performance comparison between zero-shot and fine-tuned models
0 GitHub stars

Casos de Uso

Improving sales forecasting accuracy by training on historical seasonal retail data
Developing custom energy demand models based on specific regional consumption patterns
Benchmarking specialized AI models against generic baselines to justify model deployment

Acerca de

Características Principales

Automated dataset validation and Nixtla schema formatting
Real-time status monitoring for remote training jobs
Integrated error handling for API authentication and data volume requirements
Streamlined fine-tuning job submission and configuration management
Side-by-side performance comparison between zero-shot and fine-tuned models
0 GitHub stars

Casos de Uso

Improving sales forecasting accuracy by training on historical seasonal retail data
Developing custom energy demand models based on specific regional consumption patterns
Benchmarking specialized AI models against generic baselines to justify model deployment