# Key Hyperparameters for Fine-Tuning Kronos on Custom Datasets: Configuration Guide

> Master Kronos fine-tuning with our configuration guide. Discover key hyperparameters like lookback window, predict window, epochs, and learning rates for custom datasets.

- Repository: [ShiYu/Kronos](https://github.com/shiyu-coder/Kronos)
- Tags: configuration-guide
- Published: 2026-04-10

---

**The key hyperparameters for fine-tuning Kronos on custom datasets include `lookback_window` and `predict_window` for temporal context, `tokenizer_epochs` and `basemodel_epochs` for training duration, and separate learning rates for the VQ-VAE tokenizer and Transformer predictor stages.**

Kronos is a two-stage time-series foundation model from the `shiyu-coder/Kronos` repository that combines a quantized **VQ-VAE tokenizer** with a **Transformer-based predictor**. When adapting this architecture to your own CSV files, understanding