---
dataset_info:
  features:
  - name: category
    dtype: string
  - name: author
    dtype: string
  - name: book
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 2301319885
    num_examples: 4183
  download_size: 1527950083
  dataset_size: 2301319885
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
language:
- zh
---
## Dataset summary  
This dataset is designed for Traditional Chinese (zh-tw) and comprises of a collection of books from [好讀](https://www.haodoo.net)  
**Total tokens: 1.3B**  

(Tokens are calculated by tokenizer of LLaMA2)

## Usage  
```python
from datasets import load_dataset

dataset = load_dataset("benchang1110/Taiwan-book-1B", split="train")
```