metadata
dataset_info:
features:
- name: sample_id
dtype: int64
- name: dirty_audio
dtype:
audio:
sampling_rate: 44100
- name: clean_audio
dtype:
audio:
sampling_rate: 44100
- name: predicted_audio
dtype:
audio:
sampling_rate: 44100
- name: dirty_tokens
dtype: string
- name: clean_tokens
dtype: string
- name: predicted_tokens
dtype: string
- name: noise_type
dtype: string
- name: cross_entropy_loss
dtype: float64
- name: scoreq_metrics
struct:
- name: predicted_scoreq_nr
dtype: float64
- name: predicted_scoreq_ref
dtype: float64
- name: clean_scoreq_nr
dtype: float64
splits:
- name: train
num_bytes: 106106750
num_examples: 132
download_size: 95114554
dataset_size: 106106750
configs:
- config_name: default
data_files:
- split: train
path: data/train-*