dataset_info:
features:
- name: conversation_hash
dtype: string - name: conversation
list:- name: content
dtype: string - name: role
dtype: string - name: toxicities
struct:- name: identity_attack
dtype: float32 - name: insult
dtype: float32 - name: obscene
dtype: float32 - name: severe_toxicity
dtype: float32 - name: sexual_explicit
dtype: float32 - name: threat
dtype: float32 - name: toxicity
dtype: float32
- name: identity_attack
- name: content
- name: provenance_dataset
dtype: string - name: type
dtype: string - name: timestamp
dtype: timestamp[ns] - name: language
dtype: string
splits: - name: train
num_bytes: 10619207318.606771
num_examples: 2220853
download_size: 5549604602
dataset_size: 10619207318.606771
configs: - config_name: default
data_files:- split: train
path: data/train-*
- split: train