---
title: KronosMarket
canonical_url: "https://www.modelscope.cn/datasets/lkevincc/KronosMarket"
md_url: "https://www.modelscope.cn/datasets/lkevincc/KronosMarket.md"
repository: lkevincc/KronosMarket
last_updated: 2026-10-13
license: "Apache License 2.0"
storage_size: "104 GB"
downloads: 48949
stars: 0
---

# KronosMarket

> KronosMarket - lkevincc 在 ModelScope 开源的数据集。Training data used to fine-tune Kronos, an open K-line (candlestick) prediction model.

lkevincc/KronosMarket 是 ModelScope 魔搭社区上的数据集，存储大小 104 GB，采用 Apache License 2.0 许可。

- **Repository**: lkevincc/KronosMarket
- **License**: Apache License 2.0
- **Storage size**: 104 GB
- **Downloads**: 48949
- **Stars**: 0
- **Last updated**: 2026-10-13

Source: https://www.modelscope.cn/datasets/lkevincc/KronosMarket

---

# KronosMarket

Training data used to fine-tune [Kronos](https://github.com/shiyu-coder/Kronos),
an open K-line (candlestick) prediction model.

About 111.6 GB, mostly one Parquet file per partition.

## Layout

| path | contents |
|---|---|
| `raw/OHLCV-1m/data/` | one Parquet file per month: `ohlcv_YYYY-MM.parquet` |
| `silver/binance_klines_1m/` | Binance 1-minute klines, partitioned |
| `silver/stockchina_minute_1m/` | China A-share minute bars, partitioned |
| `_smoke_test/` | small sample used to check a loader end to end |
| `dataset_infos.json` | column schema |

## raw/OHLCV-1m

Monthly Parquet files, starting from 1992-01.

| column | |
|---|---|
| `timestamp` | bar time |
| `open` `high` `low` `close` | OHLC |
| `volume` | |
| `ticker` | instrument identifier |

The same schema is declared in `dataset_infos.json`.

## silver/binance_klines_1m

Hive-partitioned by market, symbol, year and month:

```
silver/binance_klines_1m/market={spot,futures_um}/symbol=<SYMBOL>/year=YYYY/month=MM/
```

`market` separates spot from USD-margined perpetual futures. 1-minute klines run
from 2018 onward.

## silver/stockchina_minute_1m

China A-share 1-minute bars, partitioned by symbol and month. Symbols carry their
exchange suffix:

```
silver/stockchina_minute_1m/symbol=<CODE>.XSHE/year=YYYY/month=MM/
```

## Loading

```python
import pandas as pd

# one month of the raw OHLCV series
df = pd.read_parquet("raw/OHLCV-1m/data/ohlcv_2024-01.parquet")

# one partition of the Binance data
df = pd.read_parquet(
    "silver/binance_klines_1m/market=futures_um/symbol=LINKUSDT/year=2022/month=01"
)
```

The `silver/` trees are ordinary Hive partitions, so any Hive-aware reader
(Polars, DuckDB, Spark) can scan them directly.

## License

Apache License 2.0