metadata
dataset_info:
features:
- name: system
dtype: string
- name: user
dtype: string
- name: content
dtype: string
- name: prompt_tokens
dtype: int64
- name: completion_tokens
dtype: int64
- name: group
dtype: string
- name: is_toxic
dtype: int64
- name: generation_method
dtype: string
splits:
- name: train
num_bytes: 29223700
num_examples: 12000
download_size: 873726
dataset_size: 29223700
configs:
- config_name: default
data_files:
- split: train
path: data/train-*