// Fine-tunes TimeGPT on custom datasets to improve forecasting accuracy. Use when TimeGPT's zero-shot performance is insufficient or domain-specific accuracy is needed. Trigger with "finetune TimeGPT", "train TimeGPT", "adapt TimeGPT".
Generate and configure Nixtla Skills using the CLI for forecasting workflows. Use when installing or updating skills. Trigger with 'install nixtla skills' or 'update nixtla'.
Generate benchmarking pipelines to compare forecasting models and summarize accuracy/speed trade-offs. Use when evaluating TimeGPT vs StatsForecast/MLForecast/NeuralForecast on a dataset. Trigger with "benchmark models", "compare TimeGPT vs StatsForecast", or "model selection".
Research and summarize Nixtla ecosystem updates and time-series forecasting content from the web and GitHub. Use when gathering release notes, recent changes, or best-practice references. Trigger with "Nixtla updates", "what's new with TimeGPT", or "find time-series papers".
Generate production-ready TimeGPT forecasting pipeline code from requirements. Use when scaffolding a pipeline with validation, logging, visualization, and repeatable runs. Trigger with "create TimeGPT pipeline", "build TimeGPT integration", or "generate forecast code".
Analyze Nixtla baseline forecasting results (sMAPE/MASE on M4 or other benchmark datasets). Use when the user asks about baseline performance, model comparisons, or metric interpretation for Nixtla time-series experiments. Trigger with "baseline review", "interpret sMAPE/MASE", or "compare AutoETS vs AutoTheta".
Orchestrate 6-phase changelog generation workflow with AI synthesis, multi-source data fetching (GitHub/Slack/Git), quality validation, and automated PR creation. Use when automating release notes, weekly changelogs, or documentation updates. Trigger with "generate changelog", "weekly changelog", or "automate release notes".
| name | nixtla-timegpt-finetune-lab |
| description | Fine-tunes TimeGPT on custom datasets to improve forecasting accuracy. Use when TimeGPT's zero-shot performance is insufficient or domain-specific accuracy is needed. Trigger with "finetune TimeGPT", "train TimeGPT", "adapt TimeGPT". |
| allowed-tools | Read,Write,Bash,Glob,Grep |
| version | 1.0.0 |
Adapts the TimeGPT model to specific datasets for enhanced forecasting performance.
Improves forecasting accuracy by fine-tuning the pre-trained TimeGPT model on custom time series data.
This skill guides users through the process of fine-tuning TimeGPT on their own datasets. It handles data preprocessing, training configuration, and evaluation. Use when domain-specific accuracy is crucial or when the general TimeGPT model underperforms. It outputs a fine-tuned model and performance metrics.
Tools: Read, Write, Bash, Glob, Grep
Environment: NIXTLA_TIMEGPT_API_KEY
Packages:
pip install nixtla pandas scikit-learn matplotlib
Load and preprocess time series data into Nixtla format (unique_id, ds, y).
python {baseDir}/scripts/prepare_data.py \
--input data.csv \
--train_output train_data.csv \
--val_output val_data.csv
Requirements:
Output:
Create or validate training configuration.
# Create default config
python {baseDir}/scripts/configure_training.py \
--create_default \
--output config.json
# Or validate existing config
python {baseDir}/scripts/configure_training.py \
--config config.json
Configuration parameters:
Run the fine-tuning process using prepared data and configuration.
export NIXTLA_TIMEGPT_API_KEY=your_api_key
python {baseDir}/scripts/finetune_model.py \
--train_data train_data.csv \
--config config.json \
--output finetuned_model.pkl
Output: finetuned_model.pkl (serialized fine-tuned model)
Evaluate the fine-tuned model on validation data and save metrics.
python {baseDir}/scripts/evaluate_model.py \
--val_data val_data.csv \
--model finetuned_model.pkl \
--config config.json \
--output metrics.json
Output: metrics.json (MAE, RMSE metrics)
Error: NIXTLA_TIMEGPT_API_KEY not set
Solution: export NIXTLA_TIMEGPT_API_KEY=your_api_key
Error: Invalid data format
Solution: Ensure data has columns: unique_id, ds, y
Error: Insufficient training data
Solution: Provide a larger training dataset (at least 50 time series points per series)
Error: Fine-tuning failed
Solution: Adjust the fine-tuning parameters (learning rate, epochs) in config.json
Error: Missing required column
Solution: Verify CSV has unique_id, ds, y columns with correct data types
Input:
unique_id,ds,y
store_1,2023-01-01,100
store_1,2023-01-02,110
Command:
python {baseDir}/scripts/finetune_model.py \
--train_data train_data.csv \
--config config.json \
--output finetuned_model.pkl
Output: finetuned_model.pkl (a serialized, fine-tuned TimeGPT model)
Input:
unique_id,ds,y
building_1,2023-01-01 00:00,50
building_1,2023-01-01 01:00,55
Command:
python {baseDir}/scripts/evaluate_model.py \
--val_data val_data.csv \
--model finetuned_model.pkl \
--config config.json \
--output metrics.json
Output: metrics.json (performance metrics of the fine-tuned model)
{baseDir}/scripts/prepare_data.py{baseDir}/scripts/configure_training.py{baseDir}/scripts/finetune_model.py{baseDir}/scripts/evaluate_model.py