metadata

dataset_info:
  features:
    - name: query_id
      dtype: string
    - name: query
      dtype: string
    - name: positive_passages
      list:
        - name: docid
          dtype: string
        - name: text
          dtype: string
        - name: title
          dtype: string
    - name: negative_passages
      list:
        - name: docid
          dtype: string
        - name: text
          dtype: string
        - name: title
          dtype: string
    - name: subset
      dtype: string
  splits:
    - name: train
      num_bytes: 101651201606
      num_examples: 1602667
  download_size: 57281610524
  dataset_size: 101651201606
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - text-retrieval
size_categories:
  - 10M<n<100M

BGE Training Dataset (Only Retrieval Datasets)

This is a ported version of the original cfli/bge-full-data into Tevatron format containing the following 15 retrieval splits on HF:

# Splits used for training
'sts',
'msmarco_passage',
'hotpotqa',
'msmarco_document',
'nli',
'eli5',
'squad',
'fiqa',
'nq',
'arguana',
'trivial',
'fever',
'quora',
'stack_overflow_dup_questions',
'scidocsrr'

Note (Postprocessing Updates)

We pushed the whole document available in the original dataset into the text field, so the title field is empty.
The original document or query IDs were unavailable, so we created a unique query and document ID by computing the md5 hash of the text.

import hashlib

def get_md5_hash(text):
  """Calculates the MD5 hash of a given string.

  Args:
    text: The string to hash.

  Returns:
    The MD5 hash of the string as a hexadecimal string.
  """

  text_bytes = text.encode('utf-8')  # Encode the string to bytes
  md5_hash = hashlib.md5(text_bytes).hexdigest()
  return md5_hash

Please refer to cfli/bge-full-data for details and the License.