---
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.