metadata
dataset_info:
features:
- name: text
dtype: string
- name: lang
dtype: string
splits:
- name: train
num_bytes: 6316909520
num_examples: 850000
download_size: 2483196611
dataset_size: 6316909520
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Corpus for training Tokenizer
Composition : (Approximate)
Indic language:
- Bengali 36_000_000 Chars
- Tamil 30_000_000 Chars
- Kannad 30_000_000 Chars
- Telegu 24_000_000 Chars
- Malyali 21_000_000 Chars
- Gujarati 21_000_000 Chars
- Punjabi 10_000_000 Chars
Hindi : 100_000_000 Chars
English : 82_000_000 Chars
Math : 30_000_000 Chars
Code : 24_000_000 Chars
Medical: 15_000_000 Chars