Datasets:
Tasks:
Feature Extraction
Modalities:
Text
Formats:
parquet
Languages:
Japanese
Size:
10K - 100K
License:
dataset_info: | |
features: | |
- name: ids | |
dtype: int64 | |
- name: text | |
dtype: string | |
splits: | |
- name: train | |
num_bytes: 94010539 | |
num_examples: 17478 | |
download_size: 57847222 | |
dataset_size: 94010539 | |
configs: | |
- config_name: default | |
data_files: | |
- split: train | |
path: data/train-* | |
license: odc-by | |
task_categories: | |
- feature-extraction | |
language: | |
- ja | |
size_categories: | |
- 10K<n<100K | |
kajuma/CC-news-2024-July-October-cleanedを元に、9月、10月のニュースのみを抜き出したデータセット。 | |
outputトークン数を1024と想定。効率よく学習するために約1000tokensに調整(使用tokenizerは`llm-jp/llm-jp-3-13b`) |