metadata
dataset_info:
features:
- name: full_prompt
dtype: string
- name: voice_field
dtype: string
- name: emotion_description
dtype: string
- name: text
dtype: string
- name: voice
dtype: string
- name: gender
dtype: string
- name: hash
dtype: string
- name: variant
dtype: int64
- name: score_CE
dtype: float64
- name: score_CU
dtype: float64
- name: score_PC
dtype: float64
- name: score_PQ
dtype: float64
- name: json_path
dtype: string
- name: audio
dtype: audio
splits:
- name: train
num_bytes: 1183844442.52
num_examples: 7585
download_size: 1072960125
dataset_size: 1183844442.52
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Worst CE from https://huggingface.co/datasets/mrfakename/emoact-dpo