metadata
dataset_info:
features:
- name: prompt_input_ids
sequence: int64
- name: chosen_input_ids
sequence: int64
- name: rejected_input_ids
sequence: int64
- name: ref_chosen_logps
dtype: float32
- name: ref_rejected_logps
dtype: float32
splits:
- name: train
num_bytes: 344506752
num_examples: 7148
download_size: 51616889
dataset_size: 344506752
configs:
- config_name: default
data_files:
- split: train
path: data/train-*