Datasets:
Tasks:
Feature Extraction
Modalities:
Text
Formats:
parquet
Languages:
Japanese
Size:
10K - 100K
License:
File size: 681 Bytes
3851323 15bbc26 3851323 e437f1b 3851323 e437f1b 3d19d71 7cf32a8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 |
---
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`) |