metadata
dataset_info:
features:
- name: query
dtype: string
- name: document
dtype: string
- name: negatives
sequence: string
- name: metadata
struct:
- name: objective
struct:
- name: paired
sequence: 'null'
- name: self
sequence: 'null'
- name: triplet
sequence:
sequence: string
splits:
- name: train
num_bytes: 31467350418
num_examples: 3000000
download_size: 11980582978
dataset_size: 31467350418
configs:
- config_name: default
data_files:
- split: train
path: data/train-*