kronos

Kronos is a foundation model for financial K-line data. It helps forecast market sequences from OHLCV inputs and is meant for market forecasting and research.

Share on XLicense: MIT

Overview

Kronos is a family of decoder-only models trained on financial candlestick sequences, or K-line data. It addresses the gap between general time-series models and the specialized structure of market data by tokenizing OHLCV inputs into discrete tokens and using an autoregressive transformer for forecasting. It is used in Python workflows to load a model, preprocess data, and generate forecasts from market sequences.

Key features

  • Pre-trained on financial K-line sequences
  • Tokenizes OHLCV inputs into hierarchical discrete tokens
  • Uses an autoregressive Transformer for forecasting
  • Available in multiple model sizes from Hugging Face

Best for

Researchers and practitioners working with market data can use Kronos when they want a model trained on candlestick sequences and need a Python workflow for forecasting from OHLCV inputs. It fits projects focused on financial time-series research or experiments rather than general-purpose time-series models.

Upstream
shiyu-coder/Kronos
Fork on GitHub
Guo-astro/kronos
Upstream stars
40k
Category
Finance, data and research
Language
Python
License
MIT
Forked
2026-10-04
Sync status
In syncLast synced 2026-10-05