metadata
dataset_info:
features:
- name: chosen_messages
list:
- name: content
dtype: string
- name: role
dtype: string
- name: rejected_messages
list:
- name: content
dtype: string
- name: role
dtype: string
- name: chosen_reward
dtype: float64
- name: rejected_reward
dtype: float64
- name: generate_config
struct:
- name: max_new_tokens
dtype: int64
- name: repetition_penalty
dtype: float64
- name: temperature
dtype: float64
- name: top_p
dtype: float64
- name: prompt
dtype: string
- name: prompt_decode
dtype: string
- name: prompt_response_decode
dtype: string
- name: response
dtype: string
- name: response_index
dtype: int64
- name: 不適切な応答
dtype: string
- name: 指示
dtype: string
- name: 適切な応答
dtype: string
splits:
- name: train
num_bytes: 91532286
num_examples: 13603
download_size: 44519471
dataset_size: 91532286
configs:
- config_name: default
data_files:
- split: train
path: data/train-*