README.md

---
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: 106113891.0
    num_examples: 132
  download_size: 95827198
  dataset_size: 106113891.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---