metadata
dataset_info:
features:
- name: answers
sequence: string
- name: input
dtype: string
- name: instruction
dtype: string
- name: preference
dtype: int64
- name: ref_logprob_chosen
dtype: float64
- name: ref_logprob_rejected
dtype: float64
splits:
- name: train
num_bytes: 28665707
num_examples: 49383
- name: validation
num_bytes: 1171961
num_examples: 2000
download_size: 15969976
dataset_size: 29837668
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*