---
dataset_info:
  features:
  - name: index
    dtype: int64
  - name: query
    dtype: string
  - name: document
    dtype: string
  - name: drop
    dtype: bool
  - name: duplicate
    dtype: int64
configs:
- config_name: fw-edu
  data_files: fw-edu-*
- config_name: fw2-arb_Arab
  data_files: fw2-arb_Arab-*
- config_name: fw2-ces_Latn
  data_files: fw2-ces_Latn-*
- config_name: fw2-cmn_Hani
  data_files: fw2-cmn_Hani-*
- config_name: fw2-dan_Latn
  data_files: fw2-dan_Latn-*
- config_name: fw2-deu_Latn
  data_files: fw2-deu_Latn-*
- config_name: fw2-ell_Grek
  data_files: fw2-ell_Grek-*
- config_name: fw2-fas_Arab
  data_files: fw2-fas_Arab-*
- config_name: fw2-fra_Latn
  data_files: fw2-fra_Latn-*
- config_name: fw2-hun_Latn
  data_files: fw2-hun_Latn-*
- config_name: fw2-ind_Latn
  data_files: fw2-ind_Latn-*
- config_name: fw2-ita_Latn
  data_files: fw2-ita_Latn-*
- config_name: fw2-jpn_Jpan
  data_files: fw2-jpn_Jpan-*
- config_name: fw2-nld_Latn
  data_files: fw2-nld_Latn-*
- config_name: fw2-pol_Latn
  data_files: fw2-pol_Latn-*
- config_name: fw2-por_Latn
  data_files: fw2-por_Latn-*
- config_name: fw2-rus_Cyrl
  data_files: fw2-rus_Cyrl-*
- config_name: fw2-spa_Latn
  data_files: fw2-spa_Latn-*
- config_name: fw2-swe_Latn
  data_files: fw2-swe_Latn-*
- config_name: fw2-tur_Latn
  data_files: fw2-tur_Latn-*
- config_name: fw2-vie_Latn
  data_files: fw2-vie_Latn-*
license: odc-by
---

# Contrastive Pretraining

Per-language query/document pairs produced by the retrieval-common-crawl pipeline.
Each config corresponds to a single language or source with identical LightOn-style schema.

## Config overview
Configs available are: fw-edu, fw2-arb_Arab, fw2-ces_Latn, fw2-cmn_Hani, fw2-dan_Latn, fw2-deu_Latn, fw2-ell_Grek, fw2-fas_Arab, fw2-fra_Latn, fw2-hun_Latn, fw2-ind_Latn, fw2-ita_Latn, fw2-jpn_Jpan, fw2-nld_Latn, fw2-pol_Latn, fw2-por_Latn, fw2-rus_Cyrl, fw2-spa_Latn, fw2-swe_Latn, fw2-tur_Latn, fw2-vie_Latn

## Loading with 🤗 Datasets

```python
from datasets import load_dataset
ds = load_dataset("orionweller/contrastive-pretraining", "fw-edu")
```

Each row contains the columns `index`, `query`, `document`, `drop`, `duplicate`.