---
dataset_info:
  features:
  - name: title
    dtype: string
  - name: article
    dtype: string
  - name: label
    dtype: string
  splits:
  - name: train
    num_bytes: 4154802344
    num_examples: 1067358
  download_size: 2755606275
  dataset_size: 4154802344
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
This is a dataset crawled from the web from various sources based on traditional Chinese.
# Data distribution  
![Data distribution](pie_chart.png)
# Token distribution  
calculated using LlamaTokenizer  
Total tokens: 2,184,375,300 tokens (2.1B)
| news | wiki | tech | cosmopedia | human | science |
|----------|----------|----------|----------|----------|----------|
| 1.2B     | 0.43B    | 0.19B    | 0.08B    | 0.2B     | 0.08B    |