metadata
language:
- fr
dataset_info:
features:
- name: context
dtype: string
- name: question
dtype: string
- name: label
dtype: int64
- name: source
dtype: string
splits:
- name: train
num_bytes: 292674100
num_examples: 234708
- name: test
num_bytes: 49719642
num_examples: 33610
download_size: 57567043
dataset_size: 342393742
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
This is a dataset for training the cross-encoder of our RAG system. It is a combination of the PIAF, FQuAD, SQuAD-French, and pandora-s-fr datasets.